Why "rms issue targeted stock

Transcription

1 Journal of Financial Economics 56 (2000) 459}483 Why "rms issue targeted stock Julia D'Souza *, John Jacob Johnson Graduate School of Management, Cornell University, 368 Sage Hall, Ithaca, New York 14853, USA College of Business, University of Colorado at Denver, USA Received 1 August 1999 Abstract We analyze market reaction to targeted stock issuances and investigate possible motives for their use. We "nd a statistically signi"cant abnormal return of 3.61% within a three-day window around the announcement of proposed targeted stock issuances, possibly attributable to greater information on targeted stock segments as well as monitoring and motivational advantages. We "nd lower tax-loss carry forwards among "rms that issue targeted stock compared to those that spin o! segments, suggesting that tax reasons motivate targeted stock use. The return and cash #ows of targeted stocks are a!ected more by their common corporate a$liation, although industry in#uences remain strong Elsevier Science S.A. All rights reserved. JEL classixcation: G32 Keywords: Tracking stock; Diversi"cation; Corporate focus; Corporate structure; Ownership structure 1. Introduction This paper focuses on the issuance of &targeted' stock, or stock that represents an interest in the earnings of one division of a diversi"ed "rm (also called * Corresponding author. Tel.: (607) ; fax: (607) address: jd48@cornell.edu (J. D'Souza) X/00/$ - see front matter 2000 Elsevier Science S.A. All rights reserved. PII: S X ( 0 0 )

2 460 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459}483 tracking stock, letter stock, or alphabet stock). To date, 14 companies have issued 37 targeted stocks that, at one time or another, have traded in public markets. Although ownership of targeted stock entitles the holder to the bene"ts of the earnings stream of a particular industry segment, the segment for which the targeted stock is issued remains legally a part of the consolidated company. It has therefore been conjectured that managers of conglomerates use targeted stock to obtain some of the bene"ts of a spino! (e.g., reduction in the diversi"cation discount, increased security analyst interest) without the attendant loss of corporate control. The "nancial press and some academics have been critical of the use of targeted stock (e.g., Reingold, 1995; Wall Street Journal, April 11, 1995, p. 7; New York Times, July 12, 1994, p. D1). The New York Times quotes Professor Bruce Greenwald of Columbia University as saying, `It is absolutely the purest form of "nancial engineering and it yields no bene"t at alla (July 12, 1994, p. D1). Proponents, however, argue that issuing targeted stock has very real advantages over spinning o! divisions or having a single stock for a conglomerate. Brian Finn of CS First Boston asserts, `Letter stock is not always appropriate, but given the right set of circumstances it's a terri"c structurea (Wall Street Journal, April 10, 1995, p. A6). Our "rst research objective is to present empirical evidence to shed some light on this debate. Is the issuance of targeted stock accompanied by an increase in equity value? We "nd a signi"cantly positive market reaction of 3.61% within a three-day window around announcements of proposed targeted stock issuances, similar to the market reaction to spino! and equity carve-out announcements. The market appears to view targeted stock issuances as value-enhancing. Our second research objective is to investigate whether issuance of targeted stock achieves its ostensible aims of obtaining an independent valuation for the industry segments of a diversi"ed "rm and stimulating greater security analyst interest. We analyze whether the segments represented by targeted stocks operate independently or whether their corporate a$liations cause their performance to track their "rms more than their industry groupings. We "nd that the contemporaneous correlation between the returns of two targeted stocks of the same "rm is, on average, signi"cantly higher than the corresponding correlation between independent stocks in the same industries. Also, the returns of a targeted stock are more highly correlated with the returns of other targeted stocks of the same "rm than with the returns of independent "rms in the same industry, even though the di!erent targeted stocks of a "rm operate in di!erent industries. The &"rm e!ect' (which is likely to arise because of shared management, services, and liabilities) appears to dominate the &industry e!ect'. However, the industry e!ect is not smaller than that exhibited by other independent "rms in the same industry. In fact, targeted stock segments appear to track their industries more than independent "rms do, possibly because they are closer than independent "rms to being &pure plays' in their respective industries.

3 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459} Our third research question focuses on why "rms that issue targeted stock opt for this organizational structure in preference to spinning o! one or more segments. We compare targeted stock "rms to spino! "rms along three dimensions: tax status, within-"rm transfers, and "nancial leverage. In contrast to targeted stock issuances, spino!s are often taxable. We "nd that signi"cantly more spino! "rms than targeted stock "rms report tax-loss carryforwards in their "nancial statements in the year prior to the organizational structure change (26.22% versus none). Tax-loss carryforwards are likely to mitigate the adverse tax consequences of spino!s. Our "ndings suggest that tax-related issues in#uence the choice between spinning o! units or issuing targeted stocks. Additionally, issuing targeted stock enables a "rm to keep di!erent segments under the same corporate umbrella and might therefore be the preferred organizational choice when there are signi"cant intersegment transfers. To investigate this possibility, we compare the ratio of intersegment to total sales for targeted stock and spino! "rms in the year prior to the organizational change. We "nd no signi"cant di!erences between the two samples. Finally, spino!s have more direct, potentially adverse consequences for bondholder wealth than the issuance of targeted stock, since the assets of the original "rm that collateralize its debt are divided in a spino!. Bondholders have on occasion "led suit following the announcement of a proposed spino! (e.g., Marriott Corp.). Financial leverage could therefore be a factor in#uencing the choice of organizational structure. However, we do not "nd signi"cant di!erences in either the ratio of debt to total assets or interest coverage between the two sets of "rms in the year prior to the organizational change. In summary, "rms appear to issue targeted stock to achieve some of the bene"ts of a spino! without the associated adverse tax consequences and loss of corporate control. Investors perceive targeted stock issuances to be valueenhancing; we document signi"cantly positive stock price reactions around announcements of proposed issuances of such stock. However, business segments represented by di!erent targeted stocks of the same "rm do not operate independently of one another, nor is there evidence of greater analyst following subsequent to the organizational change. Targeted stock issuances might nevertheless represent good news from an information perspective because of the comprehensive "nancial statements that "rms must provide for each targeted stock segment. 2. Evolution of targeted stock issuances The earliest use of the concept of targeted stocks was General Motor's introduction of an independently traded stock for Electronic Data Systems, in conjunction with GM's acquisition of EDS in EDS's previous owner, Ross Perot, had expressed concern that the performance of EDS managers would

4 462 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459}483 have little impact on undivided GM stock. The introduction of a separate tracking stock for EDS helped convince Perot to sell the "rm (although EDS was ultimately spun o! entirely by General Motors in 1996). GM used essentially the same mechanism when it bought Hughes Aircraft in It introduced an additional targeted stock (GMH) that tracked the performance of the new subsidiary. USX followed this lead by issuing targeted stock in USX had two major divisions, U.S. Steel and Marathon Oil, which operated in widely di!erent industries. USX management felt that its undivided stock was undervalued by the market. When Carl Icahn bought 13% of USX's stock, he demanded that the company spin o! its steel division to enhance shareholder value. In response, the company decided to issue targeted stock for its steel and oil divisions instead of spinning o! the steel division. The USX targeted stock issuance di!ered from the letter stock issued by GM in two respects. First, the targeted stocks provided for the relative voting rights of the two classes of shareholders to be periodically adjusted to conform to the relative market values of the two targeted stocks. GM's letter stocks had "xed voting rights which could cause distortions if the relative market values of the stocks changed. Second, the USX targeted stocks provided for proceeds to be paid to the shareholders of targeted groups in the event of a disposition. Under equivalent circumstances, letter stock is exchanged for stock of the parent company. This provision was the source of con#ict between GM and Class H (Hughes) shareholders when GM later wanted to sell some of its Hughes operations to Raytheon because GM wanted Class H shareholders to forego their right to convert Class H shares into ordinary GM shares at a 20% premium over the prevailing price. A number of other companies have since issued targeted stock. Pittston introduced separate stock for its minerals and services divisions in 1993 and then split the services stock into two targeted stocks in CMS Energy introduced a targeted stock for its Consumer Gas subsidiary. Ralston Purina issued a targeted stock for its Continental Baking division in 1993 before selling the division in U.S. West and Tele-Communications Inc. both introduced targeted stock in 1995 (U.S. West spun o!, as MediaOne, the division represented by one of its targeted stocks in 1998.) Georgia Paci"c introduced a targeted stock for its timber division in 1997, as did Circuit City for its automobile retail unit. Sprint introduced a tracking stock for its Sprint PCS unit in Fletcher Challenge of New Zealand introduced a targeted stock for its forest division in The company issued three further targeted stocks for its paper, building, and energy divisions in Several other companies, including AT&T (after its merger with Tele-Communications Inc.), DuPont, and Quantum Corp., have announced plans for targeted stock issuances. Zi!-Davis and Donaldson, Lufkin & Jenrette, both of which have internet operations, have announced targeted stock that will track their internet interests. Other companies have attempted but failed to issue

5 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459} targeted stock. In 1993, RJR Nabisco's shareholders approved the issuance of a targeted stock for the Nabisco Food business. However, the targeted stocks were never issued, possibly because the owners of this targeted stock would not be insulated from tobacco litigation. Kmart, in 1994, tried to introduce a tracking stock for its specialty stores but, in the face of shareholder opposition, had to sell the stores instead. In August 1995, MCI announced its intention to create targeted stock to separate its long distance business from its other investments. MCI called o! its plan in September 1995, claiming that the timing was not right. 3. Characteristics of targeted stocks Targeted stock is a class of a diversi"ed company's common stock linked to the performance of a particular business unit. A company can have two or more targeted stocks. Targeted stock does not represent direct ownership interest in the targeted business, but rather an ownership interest in the entire company. Holders of a targeted stock are generally entitled to vote on matters pertaining to the entire company. The number of votes that a targeted stockholder is entitled to can either be "xed at the time of issue of the targeted stock or #oat with the market value of the di!erent targeted stocks. The issuance of targeted stock does not entail a legal division of the company. The businesses represented by the targeted stocks remain a part of the consolidated entity and share a common board of directors. Although the "rm's assets and liabilities are attributed to the various targeted businesses for "nancial reporting purposes, legal title to the assets and responsibility for the liabilities remain with the consolidated entity. Financial statements conforming to GAAP are prepared separately for each targeted business. Holders of targeted stock of a division of a company receive "nancial statements for that division in addition to the "nancials for the company as a whole. Earnings per share and dividends are also computed separately for each targeted group. The income reported by the targeted business is the basis for the payment of dividends. A company can pay dividends to shareholders of one targeted stock and not others. For instance, Circuit City paid dividends to shareholders of Circuit City Group but not to those of CarMax Group. Dividend payout and dividend yield ratios are generally di!erent across di!erent targeted stocks of the same "rm. For "ve of the seven "rms for which there are enough observations to compare dividend yield time series across targeted stocks, we "nd that these di!erences are statistically signi"cant at the 0.05 level or better. The interests of the shareholders of the various targeted stocks do not always coincide. This is particularly the case when there are sizable corporate cost allocations or a large volume of intra-company transactions such as goods sold

6 464 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459}483 or services provided by one targeted group to another. The board of directors assumes the responsibility of ensuring that the various targeted groups transact at arm's length. Some companies that have issued targeted stock have established a separate committee of outside directors to deal with such matters Targeted stocks versus equity carve-outs Issuance of targeted stock has some similarities to equity carve-outs, which are initial public o!erings of subsidiary equity. Equity carve-outs are a source of cash to the parent "rm through a public sale of equity that has a claim on the subsidiary's assets alone. Usually, the parent retains a controlling interest. The di!erence between an equity carve-out and an issue of targeted stock is that, in the case of targeted stock, there is no parent-subsidiary relationship. In addition, issuance of targeted stock usually entails no external "nancing, so there is no dilution of the ownership interest of the original shareholders Targeted stocks versus spinows The issuance of targeted stock has some advantages over a spino! as a restructuring mechanism. First, it is tax free, since targeted stock is regarded as a class of the undivided company's stock. A spino! can only be tax free if it satis"es several fairly restrictive conditions. A company can also continue to "le a consolidated tax return after the issuance of targeted stock, thereby allowing one division's losses to o!set other divisions' pro"ts. The CFO of U.S. West's Media Group estimated that issuing targeted stock, instead of doing a spino!, was likely to save U.S. West $200 million in potential taxes over a few years (Reingold, 1995). In addition, issuance of targeted stock has no adverse implications for bondholders because the company is still legally undivided. While stockholders often prefer pure plays in a single industry, bondholders are likely to prefer the more stable earnings stream of a diversi"ed "rm. Other things being equal, the cost of capital for a "rm issuing targeted stock is likely to be lower than for a "rm that spins o! a segment. Finally, a common top management and shared corporate costs are likely to result in greater cost e$ciencies when "rms opt for the issuance of targeted stock in preference to a spino!. Aron (1991) models the tradeo! between the potentially improved managerial incentives derived from a spino! and the economies of scope that come from association with the parent "rm. The issuance of targeted stock allows a "rm to bene"t from both these potential advantages. There are several reasons that the use of targeted stock has not become more widespread. First, the ownership of targeted stock does not, in general, give the holder ownership rights to the assets of the industry segment it represents.

7 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459} Voting control of the entire company has to be obtained in order to gain control of the segment's assets. For this reason, targeted stocks might not bene"t from the same `takeover premiuma as "rms that have been spun o!. Second, it is possible for management to e!ect wealth transfers between di!erent sets of targeted shareholders because of its discretion over prices for intersegment transfers of goods and services, interest rates for intersegment loans, and use of the resources of one segment for the bene"t of another. For example, Business Week (February 9, 1998, p. 106) notes that the Timber Group of Georgia Paci"c is legally obligated to sell 80% of its products to the other group at a price that matches the average price the latter pays to other suppliers and suggests that this price is lower than the price that the Timber Group could get from external customers. 4. Research questions and empirical results 4.1. Announcement-period ewects Hite and Owers (1983), Miles and Rosenfeld (1983), and Schipper and Smith (1983) "nd signi"cantly positive stock price reactions to spino! announcements. If the issuance of targeted stock confers some of the bene"ts of a spino! on shareholders, the market should also react favorably to announcements of proposed targeted stock issuances. We test this empirically using an event study methodology. We identify the date on which news of each proposed targeted stock issuance becomes public using the Wall Street Journal index, the corresponding full text article, and 8-K SEC "lings by the company. We list the "rms that have issued targeted stock, along with the relevant dates, in Table 1. Table 1 also lists companies that have announced their intention of issuing targeted stock in the future as well as those that attempted, but failed, to issue targeted stock. We extract "nancial statement information from Standard & Poor's Compustat database, the Compustat Industry Segment Database, and from individual annual reports. Data on stock prices come from the Center for Research in Security Prices (CRSP) tapes and the Wall Street Journal. We discard six data points because the announcement of the targeted stock issuance coincides with the announcement of the acquisition of the company for which the targeted stock was to be issued, and a further 11 data points because of unavailability of data. This leaves us with a sample of 12 announcements of proposed targeted stock issuances representing ten di!erent companies. To ensure the robustness of our results, we use three di!erent event study methodologies and three event windows to estimate the market reaction to announcements of targeted stock issuances. The three windows are one, two, and three days in length, ending with the day of the announcement of the proposed targeted stock issuance in the Wall Street Journal.

8 466 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459}483 Table 1 Panel A: Companies that have issued or announced issuance of targeted stock (information current as of May 1999) Company Targeted stock Date of issue General Motors AutomotiveHH October 19, 1984 EDSHH October 19, 1984 HughesHH November 18, 1985 USXH Steel GroupHH April 15, 1991 Marathon GroupHH April 15, 1991 Delhi GroupHH September 25, 1992 Ralston PurinaH Ralston Purina GroupHH June 17, 1993 Continental BakingHH June 17, 1993 Pittston CompanyH MineralsHH July 6, 1993 ServicesHH July 6, 1993 BrinksHH January 31, 1996 BurlingtonHH January 31, 1996 Fletcher Challenge Forests DivisionHH December 12, 1993 Ordinary DivisionHH December 12, 1993 Paper DivisionHH March 25, 1996 Building DivisionHH March 25, 1996 Energy DivisionHH March 25, 1996 Seagull EnergyH Enstar Alaska Approved June 1994; not issued yet. Seagull Energy Genzyme Corp. Tissue RepairHH December 16, 1994 General DivisionHH December 16, 1994 Molecular Oncology Division November 17, 1998 CMS Energy Consumers GasHH July 21, 1995 CMS EnergyHH July 21, 1995

9 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459} Tele-Communications Inc.H TCI GroupHH August 11, 1995 Liberty MediaHH August 11, 1995 Venture GroupHH September 17, 1997 US WestH Communications GroupHH November 1, 1995 MediaVision GroupHH November 1, 1995 INCO, Ltd. INCO Ltd. Common SharesHH September 9, 1996 INCO Ltd. Class VBN SharesHH September 9, 1996 Conectiv Conectiv Common Stock March 3, 1998 Conectiv Class A Common Stock March 3, 1998 Georgia Paci"cH Georgia Paci"c GroupHH December 17, 1997 Timber GroupHH December 17, 1997 Circuit City StoresH Circuit City GroupHH February 4, 1997 CarMax GroupHH February 4, 1997 Sprint Corp. Sprint Group November 24, 1998 Sprint PCS Group November 24, 1998 AT&T Business Group Announced June ; not issued yet. Consumer Group Liberty Media Group Agouron Pharmaceuticals Inc. Agouron Division Announced August 1998; not issued following merger with Warner Lambert. Oncology Division Dupont Corp. Dupont Group Announced March 10, 1999; not issued yet. Life Sciences Division Zi!-Davis ZDNet Stock Announced March 9, 1999; not issued yet. ZD Stock

10 468 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459}483 Table 1 (continued) Panel A: Companies that have issued or announced issuance of targeted stock (information current as of May 1999) Company Targeted stock Date of issue Donaldson, Lufkin & Jenrette DLJ Stock Announced March 17, 1999; not issued yet. DLJdirect Stock Quantum Corp. DSSG Stock Announced March 26, 1999; not issued yet. HDDG Stock J.C. Penney Co. J.C. Penney Group Announced May 19, 1999; not issued yet. Eckerd Drugstore Group Panel B: Companies that attempted to issue targeted stock but failed Company Targeted stock Relevant date RJR NabiscoH Nabisco Withdrawn June 23, 1993 Kmart Specialty stores Withdrawn June 3, 1994 MCIH Long distance services Withdrawn August 2, 1995 EDS was spun-o! as an independent company by GM in Ralston Purina sold the Continental Baking Group in Pittston further divided its Services targeted stock into the Brinks Group targeted stock and the Burlington Group targeted stock in Fletcher Challenge further divided its Ordinary Division targeted stock into its Paper Division, Building Division and Energy Division targeted stocks in US West spun-o! its MediaVision Group as an independent company, MediaOne, in These targeted stocks are expected to be issued following AT&T's proposed merger with TCI. HSample of "rms used in announcement period event study. HHSample of targeted stocks used in tests reported in Tables 2 and 3.

11 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459} First, we use a traditional event study procedure described in Dodd et al. (1984) with a 200-day estimation period ending 15 days prior to the announcement of the proposed targeted stock issuance in the Wall Street Journal. The parameters from the estimation period are then used to "nd the abnormal returns of each "rm on each day of the event window. We "nd a cumulative average abnormal return of 3.61% (t-statistic"3.89) for the three-day event window, 3.67% (t-statistic"4.86) for the two-day event window, and 1.76% (t-statistic"3.29) for the one-day event window. All of these t-statistics are statistically signi"cant at the 1% level. Next, we use a nonparametric event study methodology described in Corrado (1989). The procedure involves ranking the magnitude of the abnormal return for each "rm over the estimation and event periods and testing the signi"cance of the rank of the abnormal return in the event period. The Z-statistics from this test for the three-day, two-day, and one-day event windows are, respectively, 3.08, 3.77, and 2.73, all statistically signi"cant at the 1% level. Our third methodology uses a cross-sectional test described in Pilotte (1992) to allow for possible increases in return variances during the announcement period. The t-statistics from this test for abnormal stock performance during the event period are 3.35 (signi"cant at the 1% level) for the three-day event window, 4.21 (signi"cant at the 1% level) for the two-day event window, and 2.42 (signi"cant at the 5% level) for the one-day window. These results, which are consistent across methodologies and event windows, indicate a favorable market reaction to news of a planned targeted stock issuance. Zuta (1999) and Billett and Mauer (1998) also document a positive market reaction to news of proposed targeted stock issuances. The magnitude of the cumulative average two-day abnormal stock price return (3.67%) is comparable to that documented for spino!s by Schipper and Smith (1983) of 2.84% over a two-day window. It is also comparable to the (approximately) 2% abnormal return reported for equity carve-outs by Schipper and Smith (1986) and Allen and McConnell (1998). The expected potential bene"ts of a targeted stock issuance therefore appear to be comparable in magnitude to those of spino!s and equity carve-outs. We test whether the magnitude of the abnormal returns at the announcement of targeted stock issuances is signi"cantly higher than those at spino! announcements (Schipper and Smith, 1983) and at equity carve-out announcements (Schipper and Smith, 1986; Allen and McConnell, 1998) by constructing the Bonferroni 90% joint con"dence intervals around these estimates as described in Neter et al. (1985). We "nd that the con"dence interval from our study overlaps with those of these prior studies, implying that the abnormal returns documented hereare not signi"cantly di!erent from those reported in prior studies Do targeted stock segments operate independently? One of the reasons o!ered for spino!s and the issuance of targeted stock is that the market undervalues diversi"ed "rms. A spino! remedies this de"ciency

12 470 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459}483 by granting complete independence to the industry segment. Targeted stock tries to achieve some of the same e!ect by uncoupling the earnings and dividend streams of targeted segments. We investigate whether this objective is achieved by examining whether the stock returns, earnings, cash #ows, and dividends of targeted stocks track the industry to which the targeted segments belong or whether they instead track the "rm with which the segments are a$liated. In an attempt to determine whether targeted stock segments track their industry, we also compare the operational performance and valuation of segments represented by targeted stock to those of a control sample comprising independent single-segment "rms in the same industry, The sample used for these tests represents all targeted stocks issued prior to Critics of targeted stock issuance argue that targeted segments do not enjoy the operational independence that exists between a spino! "rm and its parent. Cross-dependence is likely to arise as a consequence of shared assets and liabilities, shared corporate services, and a common top management. One implication of these commonalities is that the "nancial performance measures of di!erent targeted stocks of the same "rm could be more positively correlated than similar measures for independent "rms in the same industry or a spino! "rm and its parent. On the other hand, if the common management of the "rm permits crosssubsidization across the divisions represented by the targeted stocks, there could be a negative correlation (or a reduction in the positive correlation) between the "nancial performance measures of these segments. Decisions that could induce such a negative correlation include allocations of common costs, transfer prices for intra-company sales, and interest rates for loans from one division to another Correlation between xnancial performance metrics of targeted stocks and independent xrms in the same industry We compare the correlation between the performance of targeted stocks of a single "rm to the corresponding correlation between the performance of independent "rms operating in the same industries. If targeted stocks of the same "rm operate independently of each other, these correlations should be similar. We examine "ve "nancial performance metrics: monthly stock returns, daily stock returns, quarterly earnings, quarterly cash #ows from operations, and quarterly dividend yields. For each targeted stock, we identify independent companies operating in the same four-digit SIC industry classi"cation. When there are more than 15 independent "rms in the four-digit SIC code classi"cation, we randomly choose 15 of these using a random number generator because the number of correlations to be computed increases exponentially with the number of "rms. For each performance metric, we compute the contemporaneous correlation between the metric for targeted stocks of the same "rm. The mean and the median of these correlations are presented in Column 1 of Table 2.

13 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459} For each metric, we then compute the average correlation between independent "rms in the same four-digit SIC classi"cation codes as each pair of targeted stocks. We "rst "nd all possible pairings between "rms in the two SIC codes corresponding to each pair of targeted stocks and then average the performance metric correlation between all identi"ed pairs of independent "rms. Column 2 of Table 2 presents the mean and median values of these average correlations. If targeted stock segments of the same "rm operate as independently as do their industry-matched stand-alone counterparts, we would expect to "nd no signi"- cant di!erences between correlations of targeted stocks with each other and the correlation of independent "rms in the targeted stocks' industries with one another. We use the paired nonparametric Wilcoxon signed-rank test to test for di!erences in the computed correlations between targeted stocks of the same "rm and the corresponding average correlation between independent "rms in the same industries. Using both monthly and daily data, we "nd that the correlation between stock returns of targeted stocks of the same "rm is signi"cantly positive, as is the average of the corresponding correlation between pairs of independent "rms in the same industries as the targeted stocks. For both measures, the positive correlation between two targeted stocks of the same "rm is signi"cantly greater than the correlation between their industry-matched counterparts. Targeted stocks of the same "rm do not appear to trade independently of each other. Shared top management, services, and liabilities appear to induce greater dependence in stock returns. The correlation between contemporaneous quarterly earnings of targeted stocks of the same "rm is not statistically di!erent from zero. In addition, the contemporaneous correlation between the earnings of targeted stocks of the same "rm is not statistically di!erent from the corresponding correlation for the matched independent "rms. Shared corporate a$liation does not appear to induce dependence in the earnings of di!erent targeted stocks of the same "rm. The contemporaneous correlation between quarterly cash #ows from operations for targeted stocks of the same "rm is signi"cantly positive, as is the corresponding correlation for their industry-matched independent "rms. The di!erence between the cash #ow correlation for targeted stock segments versus their industry-matched independent counterparts is signi"cant at the 5% level (two-tailed). Cash #ows have traditionally been regarded as less subject to manipulation than earnings. The results of this test suggest that there is greater dependence between the cash #ows of two targeted stock segments of the same "rm than would have obtained had they operated as independent "rms. The contemporaneous correlation between the quarterly dividend yields of two targeted stocks of the same company is not signi"cantly di!erent from zero, nor is it signi"cantly di!erent from the correlation between the dividend yields of their industry-matched "rms. The dividend streams of targeted stocks of the same "rm, like the earnings streams, appear to be independent of one another.

14 472 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459}483 Table 2 Correlations of "nancial metrics of targeted stocks The sample includes all targeted stock issuances prior to Segments associated with targeted stocks are matched with independent "rms in the same four-digit SIC code. When there are a large number of "rms in the industry, 15 "rms are selected at random to represent the industry. Targeted stock correlations are computed from the date of issuance until December 1997 or the date when the targeted stock stopped trading, whichever is earlier. Correlations for the matched independent "rms are computed for the same time period used for the associated targeted stocks. The time period used for the computation of correlations therefore falls within the 1984}1997 range. Di!erences in the number of observations used to compute the mean (median) across "nancial performance metrics arise primarily because of missing data on COMPUSTAT and CRSP. There are also di!erences in the number of observations between Columns 1 and 2 because, in some cases, independent "rms in the same four-digit SIC code classi"cation could not be found on CRSP or COMPUSTAT. Several "rms paid no dividends during the sample period, resulting in a lower number of observations for the quarterly dividend yield metric. The "gures presented in Column 2 are the means (medians) of the average correlation between pairs of independent "rms in the same industries as each pair of targeted stocks of a "rm. Each average is computed over a maximum of 225 correlations (15 15) between pairs of independent "rms. The "gures in Column 3 are the means (medians) of the average correlation between a targeted stock and independent "rms in the same industry. Each average is computed over a maximum of 15 correlations for each targeted stock. The "gures in Column 4 are the means (medians) of the average correlation between independent "rms in the same industry as each targeted stock. Each average is computed over a maximum of 105 correlations [(15 14)/2] for each targeted stock. Variable Contemporaneous correlation between pairs of targeted stocks of the same "rm Contemporaneous correlation between independent "rms in the same industries as each pair of targeted stocks Contemporaneous correlation between each targeted stock and independent "rms in the same industry (1) (2) (3) (4) Contemporaneous correlation between independent "rms in the same industry as each targeted stock Mean No. of obs. Mean No. of Mean No. of Mean No. of obs. (median) of used to (median) of obs. used (median) of obs. used (median) used to average compute average to compute average to compute of average compute the correlations the mean correlations the mean correlations the mean correlations mean (median) (median) (median) (median) Monthly stock returns (0.330) (0.096) (0.174) (0.141)

15 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459} Daily stock returns (0.173) (0.049) (0.087) (0.062) Quarterly earnings (0.087) (0.020) (0.079) (0.043) Quarterly cash #ows from (0.565) (0.018) (0.198) (0.072) operations Quarterly dividend yield (!0.010) (!0.001) (0.147) (0.173) Di!erences in correlations between Columns 1 and 2 are signi"cant at the 5% level (two-tailed) or better in the Wilcoxon signed-rank test, indicating that correlations between pairs of targeted stocks of the same "rm are signi"cantly higher than the correlations between independent "rms in the same industries. Di!erences in correlations between Columns 1 and 3 are signi"cant at the 5% level (two-tailed) or better in the Mann-Whitney test, indicating that the correlation of a targeted stock with the other targeted stocks of the same "rm is signi"cantly higher than the mean correlation between the targeted stock and independent "rms in the same industry. Di!erences in correlations between Columns 3 and 4 are signi"cant at the 5% level (two-tailed) or better in the Wilcoxon signed-rank test, indicating that the mean correlation between a targeted stock and the independent "rms in the industry is signi"cantly higher than the mean correlation between the independent "rms themselves.

16 474 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459}483 (Several of the targeted stocks and matched "rms do not pay dividends throughout the sample period, so we cannot compute correlations in dividend yield for these stocks.) Strength of xrm inyuences relative to industry inyuences The results of the tests described above indicate that there are signi"cant commonalities in some "nancial performance measures of targeted stocks of the same "rm, induced perhaps by a common senior management team and shared assets and liabilities. To gain insight into the relative strength of "rm-induced versus industry-induced commonalities, we next examine whether a targeted stock's within-"rm correlation exceeds its correlation with its industry. As in the previous set of tests, we match each targeted stock with independent "rms in the same industry. For each performance metric, we measure the contemporaneous correlation between the performance of the targeted stock and each independent "rm in the industry and compute the average of these correlations. Column 3 of Table 2 presents the mean and median values of these average correlations between targeted stock segments and their industry-matched independent "rms. As mentioned earlier, Column 1 of Table 2 shows mean and median contemporaneous correlations between performance metrics for targeted stocks of the same "rm. The statistical signi"cance of di!erences between the within-"rm and within-industry correlations is tested using the nonparametric Mann}Whitney U-Test. The contemporaneous correlations of the monthly stock returns, daily stock returns, quarterly earnings, quarterly cash #ows from operations, and quarterly dividend yields of targeted stock segments with the corresponding variables for the industry-matched independent "rms (Column 3) are all signi"cantly positive, indicating signi"cant industry e!ects for these metrics. For both the monthly and the daily stock returns, the contemporaneous correlation between two targeted stocks of the same "rm is signi"cantly more positive than the corresponding average correlation between the targeted stock and its industrymatched independent "rms. Firm-level commonalities appear to be stronger than industry in#uences. A similar result is evidenced for quarterly cash #ows from operations; "rm e!ects dominate industry e!ects. For quarterly earnings and dividend yields, however, the correlations between targeted stocks and independent industry-matched "rms are not statistically di!erent from those between targeted stocks of the same "rm. Thus, for three of the "ve measures examined (monthly stock returns, daily stock returns, and quarterly cash #ows from operations), the "rm e!ect induced by targeted stocks operating under the same corporate umbrella dominates industry commonalities. For none of the measures is the industry e!ect signi"- cantly stronger than the "rm e!ect. Targeted stocks appear to track the "rm more than the industry.

17 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459} Next, we examine whether the correlation of the performance metrics of each targeted stock with corresponding industry measures is lower than that exhibited by independent "rms in the same industry. For each metric, we compute the average industry correlation as the mean of the corresponding correlation between all possible pairs of independent "rms in the industry. The mean and median of these average correlations for the "ve "nancial measures are presented in Column 4 of Table 2. We compare these correlations with those between the targeted stock and independent "rms in the industry. We test for the statistical signi"cance of di!erences between these two sets of correlations using the Wilcoxon signed-rank test. For both daily and monthly stock returns, we "nd that the average correlation of the targeted stock with independent "rms in their industries is higher than the average correlation between the stock returns of the independent "rms themselves. We conjecture that this could be a result of a targeted stock being closer to a &pure play' in an industry than independent "rms in the same SIC classi"cation, because independent "rms could have signi"cant involvements in other industries. There are no statistically signi"cant di!erences between Columns 3 and 4 in Table 2 for quarterly earnings, quarterly cash #ows from operations, or quarterly dividend yields. When we test for di!erences across all performance metrics jointly, we "nd that the correlation between the performance of a targeted stock and that of its industry is signi"- cantly higher than the corresponding correlation for independent "rms. There is therefore no evidence that the targeted stock structure causes the industry segments that these targeted stocks represent to be less correlated with their industry than they would otherwise be. On the contrary, targeted stock segments appear to track their respective industries more closely than do independent "rms Diwerences in selected xnancial ratios between targeted stock segments and stand-alone xrms in the same industries For all years subsequent to the issuance of targeted stock, we compare ratios that measure operational performance (return on assets and return on equity), stock performance (market return), "nancial soundness (debt to total assets and interest coverage), market valuation (price to earnings and market to book), and dividend policy (dividend payout and dividend yield) across the targeted stock segments and the control sample of independent "rms in the same industries. We compare targeted stock segments to all independent single-segment "rms in the same industry using the four-digit SIC code to de"ne industry when there are at least "ve independent "rms within a classi"cation; if this criterion is not met, we use the three-digit SIC code (and in a few cases the two-digit SIC code) to de"ne industry a$liation. For each industry and year, we compute the median of each "nancial ratio for all single-segment independent "rms in that industry. The "rst column of Table 3 reports the median ratio di!erences across the time period 1989}1997 between the targeted stock segments and indepen-

18 476 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459}483 dent single-segment "rms in the same industry. Since each targeted stock segment is represented in the sample in multiple years, these observations are not strictly independent. For comparison with a sample in which the observations are independent, the second column of Table 3 reports results for the year 1996, the year with the highest number of observations. Each targeted stock segment is represented once, at most, in this sample. Table 3 Median di!erences between selected "nancial ratios of targeted stocks relative to single segment "rms in the same industry Industry is de"ned by four-digit SIC code. If fewer than "ve independent single industry "rms are available in the four-digit SIC code classi"cation, the three-digit SIC code classi"cation (or, in a few cases, the two-digit SIC code classi"cation) is used. The number of observations for some ratios are lower because of missing data on COMPUSTAT. In addition, the number of observations for the price-to-earnings and dividend payout ratios are lower because these ratios could not be computed when earnings were negative. Ratios are computed for each year, in the sample period, that data are available on COMPUSTAT. Reported "gures represent the median di!erence, in the stated period, between the relevant ratio for the targeted stock segments and the corresponding median ratio for independent "rms in the same industry. Price-to-earnings is the ratio of stock price at "scal year end to primary earnings per share excluding extraordinary items. Market-to-book is the ratio of stock price at "scal year end to book value per share of common equity. Return on equity is computed as income before extraordinary items divided by the book value of shareholders' equity. Return on assets is computed as income before extraordinary items plus after-tax interest expense, divided by total assets. Debt to total assets is the ratio of long-term debt plus debt in current liabilities, to total assets. Interest coverage is the ratio of pretax income plus interest expense, to interest expense. Market return is computed as stock price at "scal year close plus common dividends per share minus stock price at previous year's "scal year close, divided by stock price at previous year's "scal year close. Dividend payout is de"ned as dividends on common stock divided by income before extraordinary items. Dividend yield is de"ned as dividends per share of common stock divided by stock price at "scal year close. Ratio Median di!erence over Median di!erence in 1989}1997 period "scal year 1996 Price-to-earnings Market-to-book 0.111H Return on equity 0.056HHH 0.066HH Return on assets 0.002H Debt to total assets Interest coverage Market return 0.053HH Dividend payout 0.120HHH 0.119HHH Dividend yield 0.013HHH 0.010HHH No. of observations HSigni"cant at the 10% level using the Wilcoxon signed rank test. HHSigni"cant at the 5% level using the Wilcoxon signed rank test. HHHSigni"cant at the 1% level using the Wilcoxon signed rank test.

19 J. D'Souza, J. Jacob / Journal of Financial Economics 56 (2000) 459} The most striking di!erence between targeted stock segments and their comparison "rms is in the dividend payout and dividend yield ratios. Targeted stock segments pay out 12% more of their earnings in the form of dividends than do comparison "rms. This di!erence is signi"cant at the 1% level in both the 1996 and overall samples. Similarly, the dividend yield of target stocks is 1.3% higher than that of independent single-segment "rms in the industry. We conjecture that the stability that comes from being associated with a diversi"ed "rm allows these "rms to pay out a greater proportion of earnings in the form of dividends. It is also possible that these "rms pay higher dividends to signal to skeptical market participants that their rather controversial equity structure is viable. Although targeted stock segments appear to reinvest relatively lower proportions of earnings, they are not characterized by higher levels of external "nancing via debt than their independent counterparts. However, ratios involving assets or liabilities have to be interpreted with caution for targeted stock segments. Although assets and liabilities are assigned to each targeted stock segment for "nancial reporting purposes, the corporate entity retains legal title to all assets and bears legal responsibility for all liabilities. The return on equity of targeted stock segments is on average signi"cantly higher than that of single-segment "rms in the same industry. It should be noted, however, that higher dividend payments reduce the book value of shareholder equity, biasing the return on equity upward for targeted stock segments relative to their independent counterparts, all other factors being equal. None of the other ratios analyzed di!er consistently between targeted stock segments and their comparison "rms across Columns 1 and 2 of Table 3. Interestingly, the price-to- earnings ratio of targeted stock segments is not signi"cantly lower than that of independent "rms in the same industry. Thus, there is no evidence that the earnings of targeted stock segments are undervalued relative to those of their independent single-industry peers Targeted stock issuances and changes in analyst following It has been suggested that conglomerate "rms are undervalued because security analysts "nd it di$cult to value "rms that cross industry lines. For instance, Bhushan (1989) documents that an increase in the number of business segments within a company is accompanied by a drop in analyst following. Analysts generally specialize in one industry, developing industry-speci"c expertise. This expertise becomes less of a competitive advantage when an analyst follows diversi"ed "rms. Forecasting earnings and stock prices also becomes more complex because forecasts for various divisions have to be aggregated. The issuance of targeted stock has the potential to alleviate this problem by providing pure plays in single industries. If analysts avoid stocks of diversi"ed "rms because of the complexity introduced by conglomeration, then there should be an increase in analyst following after the issuance of targeted stock.