Subsidiary debt, capital structure and internal capital markets

ARTICLE IN PRESS Journal of Financial Economics 94 (2009) 327–343

Contents lists available at ScienceDirect

Journal of Financial Economics journal homepage: www.elsevier.com/locate/jfec

Subsidiary debt, capital structure and internal capital markets$ Adam C. Kolasinski University of Washington, Michel G Foster School of Business, Seattle, WA 98195, USA

a r t i c l e in fo

abstract

Article history: Received 23 June 2008 Received in revised form 6 November 2008 Accepted 9 December 2008 Available online 21 July 2009

I study external debt issued by operating subsidiaries of diversified firms. Consistent with Kahn and Winton’s [2004. Moral hazard and optimal subsidiary structure for financial institutions. Journal of Finance 59, 2537–2575] model, where subsidiary debt mitigates asset substitution, I find firms are more likely to use subsidiary debt when their divisions vary more in risk. Consistent with subsidiary debt mitigating the free cash flow problem, I find that subsidiaries are more likely to have their own external debt when they have fewer growth options and higher cash flow than the rest of the firm. Finally, I find that subsidiary debt mitigates the ‘‘corporate socialism’’ and ‘‘poaching’’ problems modeled in theories of internal capital markets. & 2009 Elsevier B.V. All rights reserved.

JEL classification: G31 G32 L22 L25 Keywords: Capital structure Internal capital markets Empirical corporate finance Subsidiary debt

1. Introduction Instead of issuing debt at the parent level, multidivision firms sometimes separately incorporate their operating divisions as distinct legal entities, i.e., as subsidiaries, and allow them to issue their own debt.

$ I thank S.P. Kothari, Stewart Myers, and Antoinette Schoar, for their advice and guidance. I thank Cliff Smith, the referee for his helpful suggestions that greatly improved the paper. I also thank Tobias Adrian, Andres Almazan, Paul Asquith, Jack Bao, Utpal Bhattacharya, Nittai Bergman, John Chalmers, Harry DeAngelo, Linda DeAngelo, Alex Edmans, Mara Faccio, Paolo Fulghieri, Ilan Guedj, Jarrad Harford, Alan Hess, Dirk Jenter, Philippe Jorion, Jiro Kondo, Andrew Lo, Roni Michaeli, Stas Nikolova, Kevin Rock, Ed Rice, Roberto Rigobon, Jeremy Stein, and Alan Timmerman as well as seminar participants at the Federal Reserve Bank of New York, George Mason University, Indiana University, MIT, UC Irvine, UC San Diego, USC, the University of Washington, and Vanderbilt University for their helpful comments and suggestions. I thank Solomon Samson of Standard & Poor’s for providing insights about legal and institutional details. I am grateful to State Farm Companies for their financial support. Any errors are my own. E-mail address: [email protected]

0304-405X/$ - see front matter & 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jfineco.2008.12.005

Holders of subsidiary debt have a claim on the subsidiary senior to that of the parent’s creditors. If the parent guarantees the debt, holders also have recourse to the parent should the subsidiary default.1 Otherwise, they have no recourse, unless the parent engages in some wrong doing (Thomson, 1991). Since 1995, subsidiary debt issues have accounted for approximately 13% of total US non-financial corporate public debt proceeds,2 yet this phenomenon remains till now largely unexamined in the empirical literature. This study seeks to fill this void by proposing and testing rationales, based on models developed in prior research, for why non-financial firms use subsidiary debt. In addition, it examines how the presence of subsidiary debt in the capital structure affects internal capital markets.

1 If the subsidiary is organized as an unlimited liability entity or a partnership with the parent as general partner, its debt holders have a claim on parent assets even without a guarantee. 2 I use data from the Securities Data Corporation (SDC) to calculate this estimate.

ARTICLE IN PRESS 328

A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

I estimate multinomial logistic models of the probability that a division is separately incorporated with its own subsidiary debt and whether that debt is guaranteed by the parent. I find that a division is more likely to have nonguaranteed subsidiary debt outstanding if it is in a firm whose divisions vary more in operating risk. This result supports the Kahn and Winton (2004) model, which predicts that firms with large cross-divisional variation in operating risk use nonguaranteed subsidiary debt to mitigate asset substitution problems. I also find some evidence that nonguaranteed subsidiary debt is more common in divisions with high cash flows and few growth options relative to the rest of the firm, suggesting firms use subsidiary debt to control Jensen’s (1986) free cash flow problem in a manner that also controls the underinvestment problem. Finally, I find that the better a division’s investment opportunities relative to the rest of the firm, the more likely it is to have parent-guaranteed subsidiary debt outstanding. Furthermore, when a division has parent-guaranteed debt outstanding, its cash flows are less likely to be diverted for investment in other divisions. As I discuss in more detail in the next section, the last two results imply that firms use parentguaranteed subsidiary debt to protect their growth divisions from the ‘‘corporate socialism’’ and ‘‘poaching’’ problems of internal capital markets modeled in Scharfstein and Stein (2000) and Rajan, Servaes, and Zingales (2000). Nonguaranteed subsidiary debt shares with project finance debt the feature of creditors having no recourse to the parent.3 I thus consider models of non-recourse project debt (John and John, 1991; Leland, 2007), in which the non-recourse feature drives the empirical predictions. I test these predictions on my sample of nonguaranteed subsidiary debt and fail to find supporting evidence. This study contributes to the large literature on debt contract features used to mitigate bondholder–shareholder conflicts and agency problems. The relevant features considered here include the specific set of assets and cash flows over which debt has priority and recourse provisions. Other examples of literature in the broad area of debt contract features include, but are not limited to, Barclay and Smith (1995a), Childes, Mauer, and Ott (2005), Guedes and Opler (1996), and Johnson (2003), who study maturity choice; Barclay and Smith (1995b), who study to priority structure; Mian and Smith (1992), Morellec (2001), and Stulz and Johnson (1985), who study security provisions; and Billet, King, and Mauer (2007), Chava and Roberts (2008), and Smith and Warner (1979), who study covenants. My results also contribute to the literature on corporate diversification and internal capital markets. Thus far, the literature has largely focused on the question of whether internal capital markets within diversified firms are efficient on average, and the evidence is mixed. Numerous studies find evidence that investment polices of diversified firms appear less efficient than those of 3 The main difference is that subsidiary debt is issued by a corporate going concern, the subsidiary, whereas project debt is issued by a special purpose vehicle created to undertake a specific, finite-lived project (Kensinger and Martin, 1988).

standalones, or that diversified firms are valued at a discount relative to standalones.4 Others, however, cast doubt on these results.5 My results imply that the issues surrounding internal capital markets are more complex. Previous research provides mixed evidence that diversification ceribus paribus reduces investment efficiency, but my findings suggest that firms take measures to mitigate this problem. Scharfstein (1998) provides evidence that improved governance can improve internal capital markets. My findings show financing policy is also important. Hence, together with Scharfstein, this study suggests that the question of how firms mitigate internal capital markets inefficiencies is as important an area of inquiry as the question of whether such markets are on average efficient. The rest of this study proceeds as follows. In Section 2, I develop rationales for subsidiary debt use and their empirical implications. In Section 3, I discuss my sample. In Section 4, I discuss my tests and present results. In Section 5, I examine alternative rationales for subsidiary debt and rule them out. Section 6 concludes.

2. Hypothesis development In this section, I use models developed in prior research to explore rationales for subsidiary debt, and I develop empirical hypotheses, summarized in Table 1.

2.1. Subsidiary debt in asset substitution models Jensen and Meckling (1976) postulate that when a firm nears financial distress, equity holders have an incentive to substitute more risky assets for existing ones in an attempt to ‘‘gamble for resurrection,’’ even if the new assets have negative net present value. Kahn and Winton (2004) argue that the problem is worse for diversified firms with divisions that differ in their operating risk. Should the riskier division encounter difficulties that threaten the entire firm with financial distress, a firm financed with parent debt will be tempted to increase the risk of the safer division. If the divisions are separately incorporated and financed with their own nonguaranteed subsidiary debt, however, financial distress at the riskier division should be less likely to affect operations in the safer division, since the holders of the riskier division’s debt have no claim over the safer division’s assets. Hence, Kahn and Winton postulate that firms will be more likely to use subsidiary debt financing when their divisions differ more in their operating risk. Flannery, Houston, and Venkataraman’s (1993) model has similar empirical predictions. I therefore postulate the following hypothesis. 4 See Berger and Ofek (1995), Lamont (1997), Lamont and Polk (2002), Lang and Stulz (1994), Ozbas and Selvili (2006), Rajan, Servaes, and Zingales (2000), Scharfstein (1998), Schoar (2002), and Shin and Stulz (1998). 5 See Campa and Kedia (2002), Chevalier (2004), Gomes and Livdan (2004), Matsusaka (2001), Graham, Lemmon, and Wolf (2002), Maksimovic and Phillips (2002), Villalonga (2004), and Whited (2001).

ARTICLE IN PRESS A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

Table 1 Empirical hypotheses. Purpose of subsidiary debt

Empirical hypothesis

Reduce asset substitution problems H1: Segments more likely to have nonguaranteed subsidiary debt in capital structure if parent firm has high dispersion in operating risk across segments, measured by industry cash flow volatility. Improve internal capital markets H2a: Having more valuable investment opportunities, measured by market-to-book and Sales Growth, relative to rest of firm increases likelihood a segment has parent-guaranteed subsidiary debt in capital structure. Improve internal capital markets H2b: Segments with parentguaranteed subsidiary debt less likely to have cash flow siphoned off and invested in other divisions. Optimize divisional capital H3: Segments more likely to have structures, trading off against cononguaranteed subsidiary debt in insurance capital structure if parent firm has higher dispersion in value of investment opportunities and segment’s cash flows more highly correlated with rest of firm. Control free cash flow diversion H4: Having fewer valuable while minimizing the investment opportunities relative underinvestment problem to rest of firm simultaneously with high cash flows increases a segment’s likelihood of having nonguaranteed debt in its capital structure.

H1. Divisions will be more likely to have subsidiary debt in their capital structure if they are in firms that have greater cross-divisional variation in operating risk. 2.2. Subsidiary debt and models of internal capital markets Scharfstein and Stein (2000, henceforth S&S) and Rajan, Servaes, and Zingales (2000, henceforth RSZ) present models of internal capital allocation wherein divisions with relatively poor investment opportunities get too much capital, and divisions rich in investment opportunities get too little. A key assumption driving the results of both models is that the CEO cannot ex ante commit to an efficient action. Eliminate this assumption and investment in both models becomes efficient. I argue that subsidiary debt can improve internal capital markets by providing a commitment mechanism. I now elaborate on how subsidiary debt fits into both models. 2.2.1. Adding subsidiary debt to the Scharfstein and Stein model S&S assume that division managers face a tradeoff between rent-seeking and productive work. A division manager’s return on productive work is higher when his division has better investment opportunities. Hence, the opportunity cost of rent-seeking is lower for opportunitypoor (‘‘low-growth’’) division managers than for opportunity-rich (‘‘high-growth’’) division managers, so low-

329

growth divisions engage in rent-seeking activities, whereas high-growth division managers do not. As a result, the CEO must pay rents to managers of low-growth divisions. The latter are empire builders, so they will accept rents in the form of an increased capital budget or in cash bonuses. The CEO wants to keep excess cash in the bonus pool for himself, so he pays rents by distorting the capital budget in the low-growth division managers’ favor rather than paying cash. Anticipating this problem, investors curtail both the firm’s capital budget and bonus pool. The CEO wants as much capital as possible, and in equilibrium fails to keep excess bonus cash because investors curtail the pool. Therefore, if it were possible, the CEO would prefer to commit ex ante to invest more in the high-growth division, which he can achieve by financing the division with subsidiary debt. Provisions on use of proceeds can bind the CEO to invest the newly raised capital in the division. Covenants restricting capital transfers can bind him to keep it there. For example, nearly all subsidiary debt issues in my sample have some covenant restricting the ability of the firm to sell the subsidiary’s assets and transfer the proceeds to other divisions. Parent securities sometimes have provisions binding the CEO to invest proceeds in certain projects. However, specifying all projects liable to get underfunded is sometimes impossible. For instance, it could be known that the best projects are in a particular division, but the precise investments needed to undertake them are not known or impossible to contractually specify. In such cases, it is optimal for the firm to commit funds to the division, as subsidiary debt allows. Separately incorporating the division as a subsidiary and giving it its own debt holders reduces the CEO’s ability to renege on such a commitment. If the low-growth division is expected to continue having relatively poor investment opportunities for some time, the CEO will want to commit to not giving it the high-growth divisions’ future cash flows. In the spirit of Jensen (1986), issuing subsidiary debt against those cash flows ensures they are distributed to creditors instead of transferred to the value division. Restrictions on use of proceeds helps ensure that the present value of those cash flows is allocated to the high-growth division at time of issuance. 2.2.2. Adding subsidiary debt to the Rajan, Servaes, and Zingales model RSZ postulate that managers of low-growth divisions can ‘‘poach’’ the surplus of high-growth divisions, and the CEO cannot commit to prohibiting it. Threat of poaching, in turn, induces high-growth divisions to make ‘‘defensive’’ investments whose surplus is lower but more difficult to poach. As a second-best solution, the CEO allocates to low-growth divisions more capital than is efficient. With more capital, low-growth division managers derive greater utility from productive work than poaching, and thus choose not to poach. High-growth divisions make the poachable, high surplus investments, albeit with less capital than is efficient.

ARTICLE IN PRESS 330

A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

If it were possible, the CEO would want to commit to prohibit poaching since it would result in higher total surplus. Financing high-growth divisions with subsidiary debt allows him to make this commitment. Public subsidiary debt legally obliges the firm to delineate the high-growth division’s profits, which can reduce the ability of low-growth division managers to claim a portion as their own. The required disclosure of the subsidiary’s profits also makes poaching more visible to outsiders, giving the CEO an incentive not to allow it. In addition, debt service requirements will shield the high-growth division surplus available to poach. Restrictive covenants on dividends to the parent can also help. 2.2.3. Formal hypotheses based on RSZ and S&S Both theories imply that diversified firms will have subsidiary debt at their high-growth divisions. Furthermore, the parent should guarantee this debt in order to minimize potential underinvestment problems in these divisions. Thus, I state the following hypothesis: H2a. Divisions are more likely to be financed with parentguaranteed subsidiary debt if they are richer in investment opportunities relative to the rest of the firm. If guaranteed subsidiary debt successfully protects the surplus of a division from poaching, cash flow of the indebted division should not get diverted to other divisions, which leads to the following empirical hypothesis: H2b. When a division has parent-guaranteed subsidiary debt outstanding, its cash flow should have less effect on the quantity of investment in other divisions. This hypothesis should also be true if firms use subsidiary debt to mitigate inefficiencies modeled in S&S. In this case, the CEO uses subsidiary debt to commit to not investing in other divisions the cash flows generated by the indebted division. 2.3. Subsidiary debt and project finance models Myers (1977) analyzes the impact of mergers on optimal capital structure. He shows that merging two firms into a conglomerate with two divisions, and jointly financing them with parent-level debt, carries both benefits and drawbacks. On the one hand such a merger has the benefit of coinsurance: poor performance at just one division can be offset by good performance at the other, reducing the ex ante probability of encountering the underinvestment problem. On the other hand, if the two firms differ in the quantity and quality of their growth options, they will differ in their optimal capital structures. Merging them, however, and using parent debt-financing, forces a single capital structure on both, which will increase the ex ante probability of the underinvestment problem to suboptimal levels in the growth firm and underutilize tax shields in the other firm. If the two firms’ optimal capital structures differ sufficiently, the drawback of forcing them to share a common one outweighs the benefit of coinsurance. John (1993) extends this analysis

and derives the conditions under which it is optimal to split up a conglomerate. John and John (1991) apply the analysis to nonrecourse project finance debt, which allows each project to have its own capital structure, but prevents coinsurance. They show that it is optimal to separately finance projects when they differ in their ex ante value of followon investment options and when their cash flows are positively correlated. Leland’s (2007) model makes similar empirical predictions. If a subsidiary is a collection of projects and growth options, these empirical predictions apply to nonguaranteed subsidiary debt. Hence, I state my next hypothesis: H3. Divisions are more likely to have nonguaranteed subsidiary debt in their capital structure when they are in a firm that has a higher cross-divisional variation in growth options and when the divisions have a higher correlation in their cash flows. 2.4. Subsidiary debt and free cash flow6 Jensen (1986) argues that firms with high free cash flows should have high leverage in order to force managers to pay out cash and not waste it. Consider a firm with some mature divisions with high cash flows and few growth options (i.e., ‘‘cash cows’’) and other divisions with many growth options. Also assume that the timing of the cash cows’ cash flows has low correlation with the optimal time to exercise the growth divisions’ options. Jensen’s argument implies that such a firm should have high leverage to prevent managers from wasting free cash flow. However, high leverage increases the likelihood of the firm encountering the underinvestment problem in the growth divisions. By levering up only the cash cows with nonguaranteed subsidiary debt and leaving the growth division and parent with low leverage, the firm can reduce free cash flow waste without inducing an underinvestment problem. Should the cash cows encounter financial distress, the fact that the debt is not guaranteed will limit their distress from inducing underinvestment in the growth divisions. Thus, I state another empirical hypothesis: H4. Divisions with simultaneously high cash flow and low-growth options relative to the rest of the firm are more likely to have nonguaranteed subsidiary debt in their capital structure. 2.5. Other candidate models Other models provide a rationale for subsidiary and project debt, but they do not appear applicable in my context. Noe (1998) argues that multinational firms separately finance subsidiaries in different countries in order to take advantage of differences in legal regimes. My sample, however, consists of domestic US subsidiaries. Chemmanur and John (1996) construct a model in which 6 I thank the referee for inspiring the hypothesis developed in this section.

ARTICLE IN PRESS A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

private benefits of control and managerial expertise drive the decision of whether to finance multiple projects separately or jointly. However, I observe neither private benefits nor managerial expertise. Equity carve-outs are similar to subsidiary debt in that its holders have a claim to only a portion of a firm. Nanda (1991) hypothesizes that carve-outs reduce information asymmetry costs. Slovin and Sushka (1997) find supporting evidence, but Vijh (2002, 2006) finds contrary evidence. However, to my knowledge, there is no model showing that subsidiary debt might be useful in mitigating information asymmetry costs. At first glance, it seems that firms near distress can use subsidiary debt to escape the underinvestment problem in a manner similar to the way Stulz and Johnson (1985) postulate firms use secured debt and Mian and Smith (1992) find that firms use debt backed by receivables. The holders of secured debt used to finance a new investment have legal priority, relative to old debt, over that new investment’s assets. Thus, by financing a new investment with secured debt instead of equity or unsecured debt, a firm near distress can prevent old debt from capturing some of the value of the new investment and thereby escape the underinvestment problem. Subsidiary debt gives the subsidiary debt holders seniority over the subsidiary’s assets and cash flows relative to parent debt holders. Therefore, it seems that a firm near distress could escape the underinvestment problem by creating a new subsidiary to undertake a new investment financed with new subsidiary debt.7 In practice, however, subsidiary debt is unlikely to be effective at controlling the underinvestment problem because the seniority of subsidiary debt is often not enforced when a firm is in distress: A strong subsidiary owned by a weak parent is generally rated no higher than the parent. The key reasons:

 the ability of and incentive for a weak parent to take 

assets from the subsidiary or burden it with liabilities during financial distress and the likelihood that a parent’s bankruptcy would cause the subsidiary’s bankruptcy regardless of its standalone strengths.

Both factors argue that, in most cases, a strong subsidiary is no further from bankruptcy than its parent and thus cannot have a higher rating. Experience has shown that bankrupt industrial companies file with their subsidiaries more often than not. 7 I empirically test whether the factors that predict secured debt issuance in Stulz and Johnson (1985) model, parent proximity to distress and relative business risk of the new investment, are associated with a higher likelihood of subsidiary debt issuance. Specifically, I test whether a business segment of a diversified firm is more likely to issue new subsidiary debt when the parent has a junk credit rating and when the segment is in an industry with lower cash flow volatility than the parent. In untabulated results, I fail to find supporting evidence.

331

yDuring severe financial distress, especially prior to a bankruptcy, a weak parent may have a powerful incentive to strip the stronger subsidiary. The court can, at best, only award monetary damages after the fact. (Samson, Sprinzen, Dubois-Pelerin, and Pfeil, 2005, p. 86) 3. Sample selection, data, and descriptive statistics 3.1. Sample selection and data Subsidiary debt issues. Using Securities Data Corporation (SDC), I identify all public and 144a subsidiary bonds and debentures issued between 1985 and 2003. I exclude issues by special purpose vehicles, branches, trusts, and finance subsidiaries, leaving 3,669 debt issues. I ignore bank loans because machine-readable loan databases do not indicate whether issuers are subsidiaries. Matching issues to Compustat segments. I match subsidiaries with debt outstanding to divisions of firms, using business segments in the Compustat database to proxy for divisions. I first attempt the match using the North American Industrial Classification System (NAICS) codes during the 1990–2003 period. I use NAICS codes rather than Standard Industrial Classification (SIC) codes because I believe the former are more precise and NAICS better reflects economic reality during my sample period than do SIC codes.8 I begin my segment panel in 1990 because I need to later merge my data with the Fixed Income Securities Database (FISD), which only contains debt with maturity dates beginning in 1990. If I cannot match a subsidiary to a business segment, I attempt to match it to an operating segment. I match using the highest level of NAICS code precision possible: first I attempt a five-digit match, then a four-digit match, and finally a three-digit match. If a subsidiary fails to match any non-financial segments of the parent firm, or if it matches all such segments, I discard it. Next, I hand-check the data using Securities and Exchange Commission (SEC) filings to ensure that each subsidiary is associated with the correct segments for each year. In some instances, a subsidiary will get matched only to a subset of the segments it encompasses. In other instances, segments in a line of businesses similar to the subsidiary are erroneously matched to it. In addition, some non-subsidiary issues are mislabeled as such. I correct all such errors by hand. Indenture data. In order to get information on bond retirements, guarantees, and parent company cross-default provisions, using CUSIP, I merge the above sample with the Fixed Income Securities Database (FISD). Some issues must be matched by hand because FISD only has the latest CUSIP of an issue, whereas SDC has only the historical CUSIP. After the merge, I am left with 1,117 bonds and debentures. Assembling the treatment panel. Using issue and retirement dates from FISD, I ensure that a subsidiary bond or 8 See Census Bureau (1993) for a discussion of the shortcomings of SIC codes. See Office of Management and Budget (1994) for a discussion on how the new NAICS system provides an improvement.

ARTICLE IN PRESS 332

A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

debenture is only associated with a given segment in years that fall between the year the debt was issued and retired, inclusive. If a firm has subsidiary debt for at least one segment-year observation, then all of its non-financial segments are in the sample for all the years the firm exists in Compustat, even in years where no segments have debt. For each segment-year observation in this treatment sample, I then obtain sales, operating income before depreciation, depreciation, and capital expenditures, as well as beginning-of-period assets and lagged sales. If the beginning-of-period assets are unavailable, I estimate them by subtracting capital expenditures from end-of-period assets and adding depreciation (the results do not change if I delete these observations). I define salesgrowth as the percentage change in current year sales over the previous year’s. I define a segment’s return on assets (ROA) as the ratio of operating income before depreciation to beginningof-year segment assets. I also obtain the parent’s property, plant, and equipment (PPE), credit rating, assets, and beginning-of-year market value of equity. I require each of the above variables in order to keep a segment-year in my sample. After applying the above sample criteria, I am left with 2,044 segment-year observations. I label this as my treatment sample, which includes 121 firms. The control panel. For every firm-year in my treatment sample, I identify the ten multisegment firms that never had any subsidiary debt, have parent-level public debt, are closest in market capitalization to the treatment firm in that year, and are not matched with another treatment firm-year observation. Of those ten, I keep in my control sample the two firms closest in tangible assets (PPE/assets) to the treatment firm. I match control firms with a treatment firm even in years where the latter has no subsidiary debt. I obtain the same variables for the control sample as I do for the treatment sample. I do not match using industry because I use industry variables in my tests, and I need variation in these variables for statistical power. I match on size and tangible assets because both these variables are related to debt capacity. Proxies for segment investment opportunities. I use two proxies: a segment’s sales growth, defined as the segment’s change in sales over the previous year normalized by the previous year’s sales, and the beginning-ofyear value-weighted market-to-book of all standalone firms in Compustat that have the same three-digit NAICS code as the segment. I define market-to-book (mkttobk) as the ratio of the market value of equity plus book liabilities (enterprise value) to total book assets. The results do not change materially if I define mkttobk as the ratio of the market value of equity to book equity or if I use the median mkttobk of all firms in the industry. Segment operating risk. The most obvious candidate to proxy for segment operating risk is segment operating cash flow volatility. Unfortunately, while it is theoretically possible to estimate it using segment operating data, such an estimate would be necessarily noisy due to the small number of data points available. I therefore use an industry estimate as a proxy. For each standalone firm in each three-digit NAICS industry, I compute earnings before taxes, interest, depreciation, and amortization (EBITDA) for every calendar quarter and normalize by

beginning-of-quarter total assets. Next, for each year, over a trailing five-year window, I estimate the standard deviation of each firm’s quarterly EBITDA to assets ratio. Then, for each industry in each year, I take the average of these standard deviations, weighting by each firm’s beginning-of-period total assets. For every industry, I am left with an annual time series of average five-year trailing quarterly cash flow volatilities. I impute to each segmentyear observation the cash flow volatility of its industry peers. I label this variable as cfvol. I use an estimate of cash flow volatility, rather than stock price volatility, because the former is a better measure of the risk of assets in place, i.e., operating risk, which is most relevant to the Kahn and Winton model. Stock price volatility also reflects the volatility of growth options. Cross-segment cash flow correlations. For each calendar quarter and industry, I compute the average of the EBITDA/Assets ratio for all standalone firms, weighting by beginning-of-period assets. Then using rolling five-year windows, for every year I compute the correlation of the EBITDA/Asset ratio of every industry pair. Next, I assume that each segment pair within each firm in a given year has the same cash flow correlation as that of their industries. I then compute the average correlation of each segment with every other segment within in the firm, weighting by the other segments’ assets, and use this figure as my proxy for a segment’s correlation with the rest of the firm. I label this variable as cf_correlation. Appending the control panel to the treatment panel, I am left with a data set of 6,250 segment-year observations. I define a trinomial categorical variable, wd, to indicate whether a segment has subsidiary debt and whether it has a parent guarantee. Following the practice of Standard & Poor’s, I treat parent guarantees and contemporaneous parent public debt with cross-default provisions identically, and henceforth refer to them collectively as ‘‘guarantees.’’ Parent cross-default provisions trigger default on parent debt should a subsidiary default, providing an incentive for a parent to bail out a distressed subsidiary, hence creating an implicit guarantee (Samson, Sprinzen, Dubois-Pelerin, and Pfeil, 2005). Proxies for debt capacity and other control variables. The standard proxies for debt capacity in the corporate finance literature typically include size, profitability, investment opportunities, operating risk, and asset tangibility. I include proxies for each at the segment level where possible, and the firm level where not. For size, I use the beginning-of-year assets of the segment. Segments do not have market capitalization, so I also include the firm’s and label it as size. For profitability, I include segment ROA. I include segment salesgrowth and industry mkttobk to proxy for its investment opportunities. I use the ratio of segment depreciation to beginning-of-year assets to proxy for asset tangibility. Since this last proxy is likely noisy, I also use the ratio of the parent’s PPE to total beginning-ofyear assets (firm_ppe/assets). As stated above, a segment’s imputed cash flow volatility, cfvol, proxies for operating risk. In addition, some segments are in industries where debt is used sparingly. To control for this effect, I include a dummy, lowlvg, which indicates the segment is in the lowest leverage quartile industry in a particular year.

ARTICLE IN PRESS A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

Finally, Wulf (2009) finds that managers of divisions that are large relative to the firm have greater influence within the firm. Hence, I use the relative size of the division, relsize, defined as the ratio of segment assets to firm assets, as a control variable. Within-firm dispersion in investment opportunities and operating risk. As my proxy for within-firm dispersion in investment opportunities (i.e., growth options), I use the intra-firm, cross-segment asset-weighted standard deviation of sales growth and mkttobk, labeled stddev (salesgrowth) and stddev(mkttobk), respectively. As a proxy for within-firm variation in divisional operating risk, I use the asset-weighted standard deviation of cfvol across segments within the firm, labeled stddev(cfvol). Differences in investment opportunities: I define variables mkttobk_difference and salesgrowth_difference as the difference between the segment’s mkttobk and salesgrowth, respectively, and the beginning-of-year assetweighted average for all other segments in the firm.

3.2. Descriptive statistics Table 3 presents descriptive statistics on firms with and without subsidiary debt outstanding for at least one segment. Variable definitions are in Table 2. Firms with subsidiary debt tend to have more tangible assets (firm_ppe/assets) and lower mkttobk. Firms with subsidiary debt also tend to be larger, consistent with prior evidence that firm size is correlated with public debt issuance due to economies of scale (Barclay and Smith, 1995a, b). Firms with subsidiary debt tend to have more cross-segment variation in salesgrowth than those without (Table 3). Table 4 presents descriptive statistics on segments with and without subsidiary debt. Segments without subsidiary debt tend to have ROA no different than other segments on average. Segments with subsidiary debt tend to have more assets than segments without, and segments without tend to represent a lower proportion of their parent firm’s assets than segments with subsidiary debt. The variable, relsize, which represents the ratio of segment to firm assets, equals 43%, 46%, and 29%, respectively, for segments with guaranteed subsidiary debt, segments with nonguaranteed subsidiary debt, and segments without debt. Like firms with subsidiary debt, segments with subsidiary debt tend to have lower mkttobk than segments without. However, segments with guaranteed subsidiary debt tend to have higher sales growth. In addition, segments with guaranteed subsidiary debt tend to have higher sales growth relative to other segments within the same firm. Utility segments appear much more likely to have subsidiary debt, consistent with prior evidence that utilities have higher debt capacity (Smith and Watts, 1992). A 36% and 46%, respectively, of segments with nonguaranteed and guaranteed debt are utilities, as opposed to 10% of segments without debt. Therefore, I include a utility dummy in my tests to control for possible effects of regulation. However, I keep utilities in my sample because of their large number and it is plausible that the factors influencing non-utilities to have subsidi-

333

Table 2 Variable definitions. mkttobk

Asset-weighted average of (market equity+book liabilities)/(book assets) for the previous period of all standalone firms with the same three-digit NAICS code as a business segment. Firm mkttobk is the asset-weighted average segment mkttobk. mkttobk_difference Difference between a segment’s mkttobk and the asset-weighted average mkttobk for all other segments in the same firm for a particular year. stddev(mkttobk) The standard deviation of mkttobk across a firm’s segments. salesgrowth Percentage change in segment sales in the current period relative to the previous period. Firm-level sales growth is the asset-weighted average of segment salesgrowth. salesgrowth_difference Difference between segment sales growth and the asset-weighted average sales growth for all other segments in the same firm for a particular year. stddev(salesgrowth) The standard deviation of sales growth across a firm’s segments in a given year. cfvol Proxy for segment cash flow volatility. Defined as the asset-weighted average of the time series standard deviations of quarterly EBITDA/assets for all standalone firms in a segment’s industry, defined by three-digit NAICS codes, over the 20 quarters prior to the year corresponding to an observation. stddev(cfvol) The standard deviation of cfvol across segments within the firm in a given year. cf_correlation The asset-weighted average of a segment’s three-digit NAICS industry cash flow correlations with the industry cash flows of all other segments within a firm. Industry cash flows defined as the asset-weighted average EBITDA/Assets. assets Beginning-of-period segment assets in $billions; firm-level assets are sum of segment assets. capx Segment capital expenditures. depreciation/assets Segment depreciation divided by assets. ppe/assets Firm property, plant, & equipment divided by beginning-of-year firm assets. lowlvg Dummy indicating that a segment belongs to a three-digit NAICS industry that is in the lowest leverage quartile of industries in a given year. relsize Relative size of a segment, defined as the ratio of a segment’s assets to the firm’s. ROA Segment operating income before depreciation divided by assets. size The firm’s beginning-of-year market capitalization in $billions. utility Dummy indicating that a segment’s primary three-digit NAICS code is 221.

ary debt do not also affect utilities. Utilities and nonutilities have comparable variation in the variables that I hypothesize to influence whether a segment has debt. Nevertheless, as a robustness check, I verify that all my results hold if I exclude utilities from my sample. 4. Empirical tests and results 4.1. Multinomial logistic regression analysis of the determinants of guaranteed and nonguaranteed debt In this section, I test my hypotheses about segment and firm characteristics that determine subsidiary debt use

ARTICLE IN PRESS 334

A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

Table 3 Descriptive statistics on firms. The sample period is from 1990 to 2003. Panel A includes firms with no subsidiary debt, and Panels B and C include firms in which subsidiary debt is not guaranteed and guaranteed by the parent, respectively. Firm sales growth and firm ROA are presented as decimals. Firm assets and firm size (market capitalization) are in $ billions. N

Mean

Std Dev

1st Pctl.

25th Pctl.

50th Pctl.

75th Pctl.

99th Pctl.

0.44 6.20 4.80 1.74 0.04 0.16 0.12 0.16 0.01 3.17

0.22 14.03 7.38 0.67 0.34 0.15 0.13 0.26 0.01 1.21

0.06 0.02 0.05 1.01 0.75 0.47 0.00 0.00 0.00 2.00

0.26 0.85 1.10 1.29 0.04 0.12 0.00 0.03 0.00 2.00

0.39 2.14 2.45 1.53 0.05 0.16 0.09 0.08 0.01 3.00

0.62 5.33 5.82 1.97 0.12 0.22 0.18 0.19 0.02 4.00

0.85 77.32 29.02 4.31 1.21 0.48 0.53 1.50 0.05 7.00

Panel B: Firms with nonguaranteed subsidiary debt firm_ppe/assets 373 0.55 firm_size 373 9.68 firm_assets 373 9.82 firm_mkttobk 373 1.50 firm_salesgrowth 373 0.05 firm ROA 373 0.15 stddev(mkttobk) 373 0.13 stddev(salesgrowth) 373 0.19 stddev(cf_vol) 373 0.02 #segments 373 3.50

0.20 19.14 13.35 0.49 0.35 0.07 0.11 0.30 0.01 1.37

0.07 0.03 0.17 0.98 0.75 0.05 0.00 0.00 0.00 2.00

0.42 0.82 1.76 1.19 0.02 0.11 0.05 0.03 0.01 3.00

0.59 2.58 4.29 1.36 0.05 0.14 0.11 0.09 0.01 3.00

0.70 8.56 12.74 1.62 0.12 0.17 0.20 0.22 0.02 4.00

0.90 95.63 66.21 3.54 1.44 0.32 0.47 1.65 0.06 8.00

Panel C: Firms with guaranteed subsidiary firm_ppe/assets 162 firm_size 162 firm_assets 162 firm_mkttobk 162 firm_salesgrowth 162 firm ROA 162 stddev(mkttobk) 162 stddev(salesgrowth) 162 stddev(cf_vol) 162 #segments 162

0.18 17.35 14.98 0.37 0.31 0.06 0.10 0.29 0.01 1.52

0.11 0.07 0.99 1.01 0.75 0.02 0.00 0.00 0.00 2.00

0.40 1.38 2.68 1.16 0.02 0.10 0.01 0.05 0.00 3.00

0.59 2.59 4.22 1.24 0.04 0.13 0.10 0.13 0.01 3.00

0.71 5.28 9.76 1.59 0.15 0.16 0.17 0.27 0.02 5.00

0.85 93.07 86.76 2.68 1.23 0.38 0.43 1.49 0.05 9.00

Panel A: Firms without subsidiary debt firm_ppe/assets 1377 firm_size 1377 firm_assets 1377 firm_mkttobk 1377 firm_salesgrowth 1377 firm ROA 1377 stddev(mkttobk) 1377 stddev(salesgrowth) 1377 stddev(cf_vol) 1377 #segments 1377

debt 0.56 7.83 9.64 1.40 0.07 0.14 0.11 0.22 0.01 3.75

(Hypotheses H1, H2a, H3, and H4). I use multinomial logistic regression analysis to model the probability of a segment having guaranteed and nonguaranteed debt. In Section 4.1.1, I describe my methods in detail. In Section 4.1.2, I describe my results (4.1.2.1) and conduct various robustness checks (4.1.2.2).

4.1.1. Research methods All my multinomial logistic specifications use a categorical dependent variable, wd, that can take on one of three values: zero if a segment has no debt, one if it has nonguaranteed subsidiary debt, and two if it has parentguaranteed subsidiary debt. The multinomial logistic regression method models the effect of explanatory variables on the probability of wd taking on each of three values, requiring that the probabilities sum to one. All specifications testing Hypotheses H1–H4 are variations on the following: Pðwd ¼ 1; 2Þ ¼ Lða þ b1 mkttobk_difference þ b2 stddevðmkttobkÞ þ b3 stddevðcf _volÞ þ b4 cf _correlation þ GControls1 Þ þ 

ðM-Model 1Þ

where controls1 ¼ omkttobk; cf _vol; ROA; lowlvg; depreciation=assets, firm_ppe; relsize; logðassetsÞ; logðsizeÞ; utility4T

and C is a coefficient row vector. All other independent variables are defined in Table 2. M-Model 2 is identical to M-Model 1, except it excludes all segments that are utilities from the sample, and hence, does not have the utility dummy. M-Model 3 is identical to M-Model 1, except it uses salesgrowth as a proxy for investment opportunities instead of mkttobk. The hypothesis that firms use subsidiary debt to mitigate asset substitution incentives, H1, implies, for nonguaranteed debt, a positive coefficient on stddev(cf_vol), a measure of cross-segment dispersion in operating risk. The hypothesis that firms use subsidiary debt to improve internal capital markets, H2a, implies, for guaranteed debt, a positive coefficient on mkttobk_difference and salesgrowth_difference, measures of how much more valuable a segment’s investment opportunities are relative to the rest of the firm. H3, the hypothesis that they use nonguaranteed debt to optimize each segment’s capital structure while giving

ARTICLE IN PRESS A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

Table 4 Descriptive statistics on segments. The sample period is 1990–2003. Panel A includes segments with no subsidiary debt, and Panels B and C include segments in which subsidiary debt is not guaranteed and guaranteed by the parent, respectively. ROA, salesgrowth and relsize are presented as decimals. Assets and are in $ billions. N

Mean Std Dev

1st Pctl.

Panel A: Segments without subsidiary debt mkttobk 5513 1.74 0.74 0.94 salesgrowth 5513 0.07 0.65 0.75 ROA 5513 0.14 0.20 0.47 cf_vol 5513 0.06 0.08 0.00 mkttobk_difference 5513 0.04 0.63 3.35 salgrowth_difference 5513 0.02 0.69 5.17 cf_correlation 5513 0.37 0.47 1.00 assets 5513 1.61 2.97 0.01 relsize 5513 0.29 0.25 0.00 depreciation/assets 5513 0.06 0.04 0.00 lowlvg 5513 0.15 0.36 0.00 utility 5513 0.10 0.31 0.00

25th Pctl.

1.23 0.08 0.08 0.01 0.17 0.15 0.00 0.20 0.09 0.03 0.00 0.00

50th 75th 99th Pctl. Pctl. Pctl.

1.52 0.04 0.15 0.03 0.00 0.00 0.34 0.62 0.20 0.05 0.00 0.00

1.98 4.48 0.14 4.44 0.23 0.85 0.08 0.56 0.25 3.37 0.12 5.15 0.84 1.00 1.67 24.30 0.41 0.99 0.07 0.26 0.00 1.00 0.00 1.00

Panel B: Segments with nonguaranteed subsidiary debt mkttobk 502 1.43 0.50 0.94 1.14 salesgrowth 502 0.04 0.42 0.75 0.03 ROA 502 0.16 0.12 0.20 0.10 cf_vol 502 0.03 0.05 0.00 0.01 mkttobk_difference 502 0.09 0.54 2.83 0.28 salgrowth_difference 502 0.06 0.56 3.79 0.14 cf_correlation 502 0.39 0.44 0.77 0.00 assets 502 4.24 5.80 0.01 0.53 relsize 502 0.46 0.29 0.01 0.21 depreciation/assets 502 0.06 0.04 0.00 0.04 lowlvg 502 0.04 0.19 0.00 0.00 Utility 502 0.36 0.48 0.00 0.00

1.23 0.04 0.14 0.01 0.05 0.01 0.36 1.69 0.43 0.05 0.00 0.00

1.54 4.48 0.12 4.44 0.19 0.85 0.04 0.29 0.03 3.23 0.10 4.46 0.87 1.00 5.29 24.30 0.71 0.98 0.06 0.26 0.00 1.00 1.00 1.00

Panel C: Segments with guaranteed mkttobk 243 1.40 salesgrowth 243 0.09 ROA 243 0.13 cf_vol 243 0.03 mkttobk_difference 243 0.03 salgrowth_difference 243 0.03 cf_correlation 243 0.52 assets 243 3.67 relsize 243 0.43 depreciation/assets 243 0.05 lowlvg 243 0.02 utility 243 0.46

1.23 0.04 0.13 0.01 0.01 0.00 0.64 2.02 0.44 0.04 0.00 0.00

1.54 4.48 0.16 4.44 0.17 0.85 0.02 0.24 0.01 2.60 0.17 4.68 0.93 1.00 4.63 24.30 0.69 0.97 0.06 0.26 0.00 1.00 1.00 1.00

subsidiary debt 0.46 0.94 1.14 0.69 0.75 0.07 0.18 0.47 0.07 0.04 0.01 0.01 0.45 1.29 0.18 0.66 2.10 0.17 0.45 0.66 0.12 5.29 0.03 0.70 0.31 0.00 0.12 0.04 0.00 0.03 0.13 0.00 0.00 0.50 0.00 0.00

up coinsurance, implies positive coefficients on stddev (mkttobk) and stddev(salesgrowth), the dispersion of costs of financial distress across segments, and cf_correlation, the correlation of a segment’s cash flows with the rest of the firm. To test whether firms lever up their cash cows with nonguaranteed subsidiary debt in order to control free cash flow wastage while minimizing the underinvestment problem (hypothesis H4), I run the following multinomial logistic model: Pðwd ¼ 1; 2Þ ¼ Lða þ b1 mkttobk_difference þ b2 stddevðmkttobkÞ þ b3 stddevðcf _volÞ þ b4 cf _correlation þ b4 ROA  mkttobk_difference þ GControls1 Þ þ

ðM-Model 4Þ

335

A negative coefficient on the interaction term for nonguaranteed debt would indicate that cash cows, that is, divisions with both relatively few growth options and high cash flow, are more likely to have nonguaranteed debt in their capital structure. I run this model only on the sample of segments with positive ROA, as the sign of the coefficient on the interaction term becomes meaningless if ROA is allowed to be negative. For robustness, I also run M-Models 5-6. Both are identical to the M-Model 4, except M-Model 5 is run on the sample that excludes utilities, and M-Model 6 uses sales growth instead of mkttobk as the proxy for segment growth options. As a further robustness check, I also run M-Model 4 through M-Model 6 using the segment’s time series mean ROA instead of realized ROA, which allows me to retain some segmentyear observations for which the realized ROA is negative but mean ROA is not. At first glance, it appears inefficient to use as the dependent variable a subsidiary debt dummy rather than some measure of the value of subsidiary debt outstanding. In treating all debt levels as equivalent, I ignore a large amount of information. This is justified, however, because theories that motivate the rationales for subsidiary debt do not have clear implications about how much debt should be placed on a subsidiary. For instance, if the CEO is using use of proceeds or disclosure requirements of subsidiary debt to commit to investing in a subsidiary or not poaching it, then the debt level need only be high enough so that debt holders demand such provisions. The lack of specific predictions about debt level makes it highly likely a debt level model will be misspecified. In contrast, the theories do provide concrete predictions about which segments should have subsidiary debt. That it is costly to retire public debt before maturity raises another concern: some segments will continue to have subsidiary debt outstanding after its benefits have ceased. Therefore, modeling the probability whether a segment issues debt, rather than whether it has debt outstanding, at first glance seems a superior research design. This approach, however, has greater drawbacks. First, it ignores the firm’s decision to keep debt at a segment rather than retire it. Twenty-four percent of the issues in my sample were retired more than a year before maturity. Therefore, the costs of early retirement are not so high as to render unimportant the decision not to retire debt. Second, CEOs are likely aware of costs of early debt retirement. Therefore, they likely issue debt at segments expected to persist in those attributes making subsidiary debt beneficial. Analyzing merely the issuance decision would thus ignore a large amount of data, reducing power. To control for time-varying macroeconomic factors, I include year fixed effects in all specifications. Previous research suggests that such factors affect firm debt policy (White, 1974; Henderson, Jegadeesh, and Weisbach, 2006). I do not use firm fixed effects because a large number of dummy variables relative to the total number of observations can lead to inconsistent estimators in non-linear models (Wooldridge, 2002, p. 484). Year fixed effects present no problems because they are few in number. Observations associated with the same firm are likely correlated with one another. Such correlation could

ARTICLE IN PRESS 336

A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

potentially bias my standard errors and hence overstate significance. I therefore compute all standard errors using the cluster method (i.e. the Rogers method), defining clusters by firm. I thus make my standard errors robust to arbitrary correlation among observations associated with the same firm. For a formal treatment of the method, see Rogers (1993), Wooldridge (2002, p. 405–407), and Williams (2000). 4.1.2. Multinomial logistic regression results 4.1.2.1. M-Models 1 and 4. The parameter estimates of MModel 1 are given in Table 5, Panel A. The results support the hypothesis that firms use nonguaranteed debt to mitigate asset substitution problems (H2). They also strongly support the hypothesis that firms use parent-guaranteed subsidiary debt to mitigate internal capital market inefficiencies (Hypothesis H2a). They fail to support the hypothesis that firms use nonguaranteed debt to allow unique capital structures for segments (H3). The results of M-Model 4 are in Panel B of Table 5, and they support the hypothesis that firms lever up their cash-cow divisions to control the free cash flow problem while minimizing the underinvestment problem (H4). Asset substitution (Hypothesis H1). The coefficient on stddev(cf_vol) is positive and significant at the 5% level for nonguaranteed debt, implying that greater dispersion of

cash flow volatility across segments within a firm increases thee likelihood of nonguaranteed subsidiary debt use. The value also has some economic significance. It implies that when all variables are at their means, a one standard deviation increase in operating risk dispersion increases the probability that a segment has nonguaranteed debt by 0.5 percentage points. While small in absolute terms, this number has some significance compared to the base probability of 5.1%. Internal capital markets (Hypothesis H2a). The coefficient on mkttobk_difference is highly positive and significant at the 1% level for guaranteed debt in Model M-1, meaning segments with mkttobk higher than the rest of their firm are more likely to have subsidiary debt outstanding, supporting Hypothesis H2a. Recall that both the S&S and RSZ theories of internal capital markets imply that firms can improve internal capital markets by issuing guaranteed subsidiary debt at divisions with the highest level of mkttobk within the firm. The coefficient estimate is also economically significant. Its value implies that when all variables are at their means, a one standard deviation increase in mkttobk_difference increases the probability that a segment has guaranteed subsidiary debt by 1.5 percentage points. While at first glance this seems like a modest amount, it is high compared to the base probability of 1.46%, the probability of a segment

Table 5 The determinants of subsidiary debt use. Parameter estimates of a multinomial logistic regression analysis of whether a segment has guaranteed or nonguaranteed debt outstanding in a given year. Rogers standard errors, clustered by firm, are in parenthesis. All regressions include year fixed effects. Variable definitions are in Table 2. Panel B excludes observations with negative ROA. The sample period is from 1990 to 2003. M-Model 1

Panel A Test variables mkttobk_difference stddev(mkttobk)

M-Model 2

M-Model 3

Non-guaranteed

Guaranteed

Non-guaranteed

Guaranteed

0.326 (0.195) 1.302 (1.152)

0.952 (0.348) 0.915 (1.576)

0.384 (0.257) 0.337 (1.066)

0.640 (0.346) 1.351 (1.737)

salesgrowth_difference stddev(salesgrowth) stddev(cf_vol) cf_correlation Control variables mkttobk

0.098 (0.046) 0.267 (0.318)

0.090 (0.059) 0.063 (0.338)

0.103 (0.049) 0.029 (0.384)

0.087 (0.065) 0.198 (0.513)

0.534 (0.241)

0.930 (0.451)

0.526 (0.235)

0.659 (0.406)

salesgrowth cf_vol ROA Assets Relsize depreciation/assets

4.654 (1.642) 1.059 (0.447) 0.071 (0.024) 1.773 (0.341) 1.113 (2.800)

6.039 (2.490) 0.269 (0.712) 0.020 (0.032) 2.011 (0.513) 1.212 (3.983)

4.858 (1.777) 0.990 (0.593) 0.088 (0.026) 2.175 (0.440) 1.333 (3.495)

5.974 (2.575) 0.722 (0.583) 0.038 (0.036) 1.451 (0.785) 0.428 (4.815)

Non-guaranteed

Guaranteed

0.094 (0.143) 0.356 (0.306) 0.107 (0.045) 0.381 (0.310)

0.457 (0.184) 0.831 (0.398) 0.094 (0.057) 0.023 (0.346)

0.299 (0.203) 5.193 (1.732) 1.070 (0.439) 0.079 (0.024) 1.628 (0.338) 1.355 (2.831)

0.536 (0.274) 6.715 (2.698) 0.229 (0.677) 0.036 (0.031) 1.856 (0.516) 0.942 (4.070)

ARTICLE IN PRESS A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

337

Table 5 (continued ) M-Model 1 Non-guaranteed firm_ppe/assets log(firm_size) Utility Lowlvg Observations Pseudo-R2

Panel B Test variables mkttobk_difference * ROA

M-Model 2 Guaranteed

Non-guaranteed

0.044 (0.938) 0.013 (0.009) 1.409 (0.552) 1.791 (0.373) 6250

1.452

1.264 (0.684) 0.003 (0.005) 1.091 (0.347) 0.833 (0.448) 6250

0.17 M-Model 4

M-Model 3 Guaranteed

Non-guaranteed

Guaranteed

(0.705) 0.003 (0.005)

0.789 (0.918) 0.011 (0.009)

1.412

0.759 (0.460) 5382

1.739 (0.358) 5382

0.370 (0.938) 0.007 (0.008) 1.596 (0.535) 1.915 (0.448) 6257

0.14 M-Model 5

0.17 M-Model 6

Non-guaranteed

Guaranteed

Non-guaranteed

Guaranteed

2.432 (0.981)

0.752 (1.065)

2.502 (1.038)

0.049 (1.153)

salesgrowth_difference * ROA Controls variables mkttobk_difference

0.727 (0.262)

1.216 (0.477)

0.813 (0.326)

stddev(mkttobk)

0.663 (0.945) 1.418 (1.503)

0.821 (0.791) 1.060 (1.170)

0.540 (0.959) 0.084 (1.087)

1.378 (1.231) 2.018 (1.649)

stddev(salesgrowth) stddev(cf_vol) cf_correlation mkttobk

0.098 (0.058) 0.179 (0.357) 1.232 (0.431)

0.078 (0.051) 0.442 (0.328) 0.484 (0.243)

0.077 (0.056) 0.243 (0.394) 0.486 (0.241)

0.092 (0.061) 0.066 (0.545) 0.885 (0.378)

salesgrowth cf_vol assets relsize depreciation/assets firm_ppe/assets log(firm_size) utility lowlvg Observations Pseudo-R2

4.275 (1.652) 0.069 (0.025) 1.733 (0.353) 4.320 (3.039) 1.023 (0.673) 0.003 (0.005) 1.230 (0.338) 0.830 (0.462) 5563

6.275 (2.718) 0.031 (0.032) 2.339 (0.541) 0.872 (4.317) 0.001 (0.969) 0.010 (0.010) 1.434 (0.583) 2.181 (0.639) 5563 0.18

Non-guaranteed

Guaranteed

0.721 (0.482)

0.611 (0.419)

0.059 (0.174) 0.510 (1.053)

0.374 (0.205) 0.019 (1.131)

0.276 (0.330) 0.103 (0.045) 0.509 (0.315)

0.648 (0.443) 0.112 (0.052) 0.208 (0.342)

0.261 (0.231) 4.977 (1.787) 0.073 (0.024) 1.639 (0.353) 3.747 (3.228) 1.156 (0.645) 0.002 (0.005) 1.331 (0.325) 0.972 (0.442) 5569

0.455 (0.320) 6.941 (2.988) 0.045 (0.031) 2.156 (0.553) 0.749 (4.475) 0.328 (0.977) 0.005 (0.009) 1.676 (0.577) 2.392 (0.710) 5569

0.710 (0.449)

salesgrowth_difference ROA

(0.655) 0.000 (0.005) 1.239 (0.338) 1.006 (0.433) 6257

4.422 (1.757) 0.087 (0.026) 2.154 (0.453) 1.310 (3.636) 1.171 (0.707) 0.004 (0.005)

6.195 (2.796) 0.037 (0.034) 1.886 (0.837) 0.560 (5.927) 0.686 (0.932) 0.010 (0.009)

0.770 (0.478) 4754

2.102 (0.633) 4754 0.15

0.18

 Significant at 10%.  Significant at 5%.  Significant at 1%.

having parent-guaranteed subsidiary debt when all variables are at their means. It is also high compared to the unconditional probability of 3.8%.

The coefficient on mkttobk_difference is also positive and marginally significant for nonguaranteed debt, having a p-value of 0.095. This result is consistent with some

ARTICLE IN PRESS 338

A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

subsidiary debt issues having implicit guarantees that I cannot observe (Stumpp, Rottino, Fanger, Stephan, and Carter, 2003). Optimizing divisional capital structures (Hypothesis H3). The coefficients on cf_correlation and stddev(mkttobk) are not statistically significant for nonguaranteed subsidiary debt. Hence, I fail to find evidence that firms use subsidiary debt to trade off between optimizing individual segment capital structures and coinsurance. Levering up cash cows (Hypothesis H4). The coefficient on the interaction term in M-Model 4 in Panel B of Table 5 is negative and significant. This result indicates that simultaneously having high cash flow and fewer growth options than the rest of the firm significantly increases the likelihood that a segment has subsidiary debt, confirming hypothesis H4. To assess economic significance, I examine odds ratios, which are more straightforward to analyze than marginal effects when there are interaction terms in a logit specification. The coefficient on ROA equals 0.821. This implies that in a segment with growth option characteristics identical to the rest of the firm (mkttobk_difference ¼ 0), a one standard deviation increase in ROA of 20 percentage points will tend to change the odds of a segment having nonguaranteed debt by a factor of 0.85 ¼ exp(0.81 * 0.2). Alternatively stated, the odds of a segment with the same growth option profile as the rest of the firm sees its odds of having nonguaranteed debt decrease by 15% when ROA moves up by one standard deviation. Now consider a segment with fewer growth options relative to the rest of the firm, its mkttobk_difference ¼ 1. Note the coefficient on the interaction term is a 2.432. If such a segment’s ROA increases by 20 percentage points, its odds of having nonguaranteed subsidiary debt in its capital structure increases by a factor of 1.38 ¼ exp[0.2 * (0.81+2.432)]. Alternatively stated, the odds of a segment with mkttobk_difference ¼ 1 having nonguaranteed subsidiary debt outstanding increases by an arguably significant 38% when its ROA increases by one standard deviation. As a robustness check, I rerun M-Models 4–6 using the segment’s time series mean ROA instead of the realized ROA in that year. By using mean ROA instead of realized ROA, I am able to retain some of the observations that were dropped because realized ROA happens to be negative for a particular segment in a given year. In untabulated results, I find the coefficients have the same signs as before and are of a qualitatively similar magnitude. However, the statistical significance of the interaction term coefficient in M-Models 4 and 5 is slightly reduced. When realized ROA is used, the interaction term p-values are, respectively, 0.01 and 0.02. When time series mean ROA is used, the p-values increase to 0.07 and 0.10. The reduced p-values are likely due to reduced statistical power that stems from ignoring meaningful time series variation in the data. Other observations: The various proxies for segment debt capacity have the predicted signs. Market-to-book and sales growth unconditionally decrease the probability of a segment having all forms of subsidiary debt, consistent with the notion that they capture some element of financial distress costs. Having more assets

and a higher ROA increases the probability of nonguaranteed debt, but not guaranteed debt. This result suggests that a parent guarantee eliminates some of the need for a segment to have good credit quality on its own to carry debt. Parent asset tangibility, measured as firm_ppe/assets, increases the likelihood of a segment having nonguaranteed debt. Since segment and parent asset tangibility are likely correlated, this result is consistent with the notion that asset tangibility increases debt capacity. Not surprisingly, utilities and segments not in low leverage industries are more likely to have subsidiary debt, as indicated by the positive and negative coefficients on utility and lowlvg, respectively. Being large increases a segment’s probability of having its own debt, of both types, consistent with prior research that finds that larger entities are more likely to issue public debt due to the scale economies involved in this sort of financing (Barclay and Smith, 1995a, b). 4.1.2.2. Robustness (M-Models 2–3 and 5–6). To ensure that utilities are not driving my results, I estimate M-Models 2 and 5, which are identical to M-Models 1 and 4, except I exclude all utilities from the sample. The results are qualitatively unchanged, as shown in column 2 of Panels A and B of Table 5. Since industry market-to-book likely measures segment investment opportunities with error, I estimate M-Model 3, which is identical to M-Model 1, except that it uses salesgrowth, instead of mkttobk, to estimate segment investment opportunities. It also uses the difference between a segment’s salesgrowth and the rest of the firm to measure the relative quality of its investment opportunities, and it uses the cross-segment standard deviation in salesgrowth to measure dispersion in costs of financial distress within the firm. The coefficient on salesgrowth_difference for guaranteed debt is positive and statistically significant in M-Model-3, but coefficients on stddev(salesgrowth) and cf_correlation are not statistically significant for nonguaranteed debt. Hence Model M-3 provides the same qualitative inferences as Model M-1. Curiously, stddev(salesgrowth) is statistically significant for guaranteed debt, which none of the hypotheses examined predict. M-Model 6 is identical to M-Model 4, except it uses salesgrowth, salesgrowth_difference, and ROA * salesgrowth_ difference in place of mkttobk, mkttobk_difference, and ROA * mkttobk_difference. The coefficient on the interaction term is no longer significant. My test of Hypothesis H4 is not as robust as my tests of Hypotheses H1–H3. To determine whether cross-sectional or time series effects are driving my results, I rerun all models, except I replace all independent variables with their time series means. The results, available upon request, are qualitatively unchanged, indicating that the effects are primarily cross-sectional. 4.2. Investment-cash flow sensitivities In this section, I examine the extent to which subsidiary debt protects a segment’s cash flows from getting diverted and invested in other segments. Section

ARTICLE IN PRESS A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

4.2.1 discusses my research design and the results. Section 4.2.2 tests whether endogeneity is driving the results. 4.2.1. Research design and results As discussed in Section 2, if subsidiary debt protects segments from poaching, then subsidiary debt in a segment’s capital structure should reduce the extent to which its cash flows are diverted and invested in other segments. Empirically, this implies that indebted segments’ cash flows have a smaller effect on investment in other segments (Hypothesis H2b). If firms are primarily using parent-guaranteed subsidiary debt for anti-poaching purposes, then I would expect a stronger effect for parent-guaranteed debt. In order to test H2b, I take every segment-year observation in my sample, and I pair it with every other segment-year observation within the firm in that year. Thus, for example, if a firm has n segments in a year, it will have n(n1) observations in my expanded sample for that year. I need to pair segments in this manner because I wish to test how the sensitivity of a segment’s investment to another segment’s cash flows changes according to whether that other segment has subsidiary debt in its capital structure. This method will cause firms with more segments to be more heavily weighted in the analysis, but this is appropriate as theory predicts that more diversified firms are more likely to have a propensity for internal capital market inefficiencies. My method likely induces cross-sectional correlation in the error term, but I address this problem below in how I compute standard errors. Next, using a specification similar to that of Shin and Stulz (1998), I estimate the following equations for this expanded panel of within-firm segment pairs: capxit ¼ a þ b1 own_cf it þ b2 other_cf it þ b3 subdebt jt þ b4 other_cf it  subdebt jt þ b5 logðsizeit Þ þ b6 junkit þ b7 Q it þ b8 salesgrowthit þ ijt 8iaj ðCF-Model 1Þ capxit ¼ a þ b1 own_cf it þ b2 other_cf it þ b3 guarsubdebt jt þ b4 other_cf it  guarsubdebt jt þ b5 logðsizeit Þ þ b6 junkit þ b7 Q it þ b8 salesgrowthit þ ijt 8iaj ðCF-Model 2Þ

339

where capx is segment i’s capital expenditures, own_cf is segment i’s own cash flow, and other_cf is the cash flow of another segment within the firm, which I label segment j. All three of these variables are normalized by segment i’s beginning-of-period assets, and cash flow is defined as operating income before depreciation and amortization. The variables subdebt and guarsubdebt are dummies indicating segment j (the other segment) has, respectively, subsidiary debt or parent-guaranteed subsidiary debt outstanding. Junk is a dummy variable indicating the firm has a junk credit rating, a proxy for capital constraints. All other variables are defined as before. I include mkttobk and salesgrowth because they are proxies for segment investment opportunities. H2b predicts a negative coefficient on the interaction term in CF-Model 2, as such a finding would indicate that cash flows of segments with parent-guaranteed debt are less likely to be diverted for investment in other segments. I cluster standard errors by firm, ensuring robustness to heteroskedasticity and arbitrary serial and within-firm cross-segment correlation in the error term. Clustering by firm thus ensures that standard errors will not be biased because some segmentyears appear multiple times in my sample. All regressions include industry fixed effects, defined by segment three-digit NAICS codes. I do not use segment fixed effects because the vast majority of segments that have subsidiary debt in at least one year also have it for all years they appear in the sample. Thus, segment dummies would have a high degree of colinearity with the subsidiary debt dummies, and so including segment fixed effects in the specification would bias my standard errors upward. In untabulated results, I confirm that when I use segment fixed effects, all my coefficients have the predicted signs but also higher standard errors. The descriptive statistics for all the variables used in the above two models are presented in Table 6. Estimation results for CF-Models 1 and 2 are in columns 1 and 2 of Table 7. Consistent with Shin and Stulz (1998), in both models a segment’s own sales growth and cash flow, along with other segments’ cash flow, all significantly affect segment investment. Strangely, market-to-book does not have an effect in either model, but this is most likely due to the

Table 6 Descriptive statistics on investment-cash flow regression sample. To assemble this sample, each segment-year observation used in previous tests is paired with every other segment within the same firm that year. Thus, if a firm has three segments in a given year, it will have six observations associated with it in that year in this sample. A firm with n segments in a given year will have (n1)n observations associated with it in that year in this sample. The variables own_cf, capx, mkttobk, and salesgrowth refer to a segment’s own cash flow (defined as operating income plus depreciation and amortization), capital expenditures, normalized by the segment’s own beginning-ofperiod assets, as well as its industry market-to-book, and sales growth. The variable other_cf corresponds to the other segment’s cash flow, normalized by the first segment’s beginning-of-period assets. Junk is a dummy variable indicating that the parent firm has a junk credit rating, and firm_size is the parent firm’s market capitalization as of the end of the previous fiscal year, in $ billions.

capx own_cf other_cf mkttobk salesgrowth junk firm_size

N

Mean

Std Dev

1st Pctl.

25th Pctl.

Median

75th Pctl.

99th Pctl.

17,081 17,081 17,081 17,081 17,081 17,081 17,081

0.098 0.154 0.652 1.679 0.067 0.146 10.320

0.134 0.496 2.766 0.688 0.672 0.353 20.187

0.000 0.529 0.267 0.935 0.746 0.000 0.017

0.030 0.077 0.029 1.201 0.096 0.000 1.170

0.062 0.143 0.138 1.454 0.032 0.000 2.886

0.110 0.223 0.441 1.930 0.147 0.000 8.597

0.894 0.871 9.204 4.476 4.434 1.000 95.634

ARTICLE IN PRESS 340

A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

Table 7 Investment-cash flow regressions. Results from an OLS regression of a segment’s own capital expenditures, normalized by beginning-of-period assets, on its own cash flow (own_cf), the cash flow of another segment (other_cf), dummies indicating whether the other segment has subsidiary debt (subdebt) or parent-guaranteed subsidiary debt (guarsubdebt), and proxies for investment opportunities (mkttobk and salesgrowth) as well as the parent firm-level variables firm_size and junk. The variables div_rest and mequity are dummies indicating whether the other segment has debt covenants restricting dividends to the parent or a minority equity stake outstanding. The sample is the same as described in Table 6. Rogers standard errors, clustered by firm, are in parentheses. Each specification includes industry and year fixed effects.

own_cf other_cf subdebt * other_cf

CF-Model 1

CF-Model 2

CF-Model 3

CF-Model 4

CF-Model 5

0.029

0.029

0.029

0.029

(0.008) 0.005 (0.001) 0.001 (0.003)

(0.008) 0.007 (0.001)

(0.008) 0.007 (0.001)

(0.008) 0.007 (0.001)

0.029 (0.008) 0.007 (0.001)

0.006 (0.002)

0.006 (0.002) 0.001 (0.003)

0.004 (0.002)

0.006 (0.002)

guarsubdebt * other_cf utility * other_cf div_rest * other_cf

0.002 (0.002)

mequity * other_cf subdebt

0.006 (0.007)

0.011 (0.006)

guarsubdebt

0.004 (0.008)

0.004 (0.008)

div_rest

0.004 (0.010) 0.005 (0.012)

mequity mkttobk salesgrowth log(firm_size) junk Observations R2

0.005 (0.007) 0.037 (0.009) 0.002 (0.002) 0.004 (0.014) 17,081 0.17

0.005 (0.007) 0.036 (0.009) 0.002 (0.002) 0.004 (0.014) 17,081 0.18

0.005 (0.007) 0.036 (0.009) 0.002 (0.002) 0.004 (0.014) 17,081 0.18

0.005 (0.007) 0.036 (0.009) 0.002 (0.002) 0.004 (0.014) 17,081 0.18

0.000 (0.009)

0.030 (0.018) 0.004 (0.007) 0.036 (0.009) 0.002 (0.002) 0.004 (0.014) 17,081 0.18

 Significant at 10%.  Significant at 5%.  Significant at 1%.

fact that market-to-book is an industry variable and I include industry dummies in the regressions. In untabulated results, where, like Shin and Stulz, I use segment fixed effects instead of industry fixed effects, the coefficient on market-to-book becomes positive and significant. The interaction of subdebt with other_cf in CF-Model 1 does not have a significant effect on investment. Hence, on average, subsidiary debt, without distinguishing whether it is parent-guaranteed, does reduce the extent to which a segment’s cash flow is related to investment in other segments. However, since not all subsidiary debt is likely designed to prevent cross-segment poaching, and nonguaranteed debt is often placed on segments with poorer investment opportunities, this result is not surprising. Hence, in CF-Model 2 I run the same test, except I interact guarsubdebt with other_cf. In this new specification, the coefficient on the interaction is statistically significant at the 1% level, and it has the predicted negative sign. The economic magnitude is also large. The estimated sign of the coefficient on the interaction term is a negative 0.006,

whereas the sign on other_cf is 0.007. Hence having parent-guaranteed subsidiary debt outstanding nearly eliminates the extent to which a segment’s cash flow is related to investment in other segments, just as Hypothesis H2b predicts. To ensure that this result is not driven by utility segments, I estimate another equation, CF-Model 3, which is identical to CF-Model 2 except that in includes the interaction other_cf with a dummy indicating that a segment is a utility (has a three-digit NAICS code equal to 221). The results, presented column 3 of Table 7, are qualitatively similar to that of CF-Model 2. A natural question to ask is what characteristic of guaranteed subsidiary debt prevents cash flows from an indebted segment from being diverted for capital expenditures in other segments. Debt covenants represent one possibility. Using public filings, I examine the covenants of all subsidiary debt issues in my sample. I define a dummy variable div_rest which equals one if segment j has a subsidiary debt issue with a covenant that places

ARTICLE IN PRESS A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

restrictions on dividend payments to the parent, and zero otherwise.9 I then estimate an equation identical to CFModel 2, except I include div_rest and its interaction with other_cf in the specification. The result is given in column 4 of Table 7. The coefficient on interaction of other_cf and guarsubdebt decreases in absolute magnitude, but it is still statistically significant while the interaction of div_rest and other_cf is of only marginal significance. The evidence that dividend restrictions drive of the anti-poaching effect of guaranteed subsidiary debt is weak. However, this could be due to low power since I only observe public subsidiary debt, which in contrast to bank debt, tends to have fewer covenants. It is also possible that some other omitted variables correlated with the presence of guaranteed debt, and not guaranteed debt itself, are driving the result in CF-Model 2. One possibility is minority equity stakes. I therefore define a dummy variable mequity that indicates whether segment j has minority equity outstanding in a given year. I then estimate an equation identical to CF-Model 2, except I include mequity and its interaction with other_cf in the new specification, which I label as CF-Model 5. The coefficient on the new interaction is not statistically significant, and the results of CF-Model 2 are qualitatively unchanged. I conclude minority equity stakes are not driving my results. The presence of minority equity stakes is only one potential omitted variable. I cannot rule out the alternative hypothesis that firms, when they use guaranteed subsidiary debt, undertake other measures to protect a subsidiary’s cash flow that I fail to observe and that are not directly tied to the debt. Since a guarantee indicates a firm is highly committed to the subsidiary, this hypothesis has some plausibility. Even if it were true, however, my results would still imply that, at the very least, guaranteed subsidiary debt often comes with a package that protects high-growth segment cash flows from diversion. In the next section, I examine and rule out the possibility that endogeneity is driving the results.

4.2.2. Ruling out endogeneity The firm’s decision to issue parent-guaranteed debt at a segment is endogenous. It is possible that firms, for reasons unrelated to poaching or corporate socialism, issue such debt at a segment when the latter’s cash flows are less correlated with the investment in other segments, providing an alternative explanation to my results. To rule out this possibility, I examine whether firms are more likely to place parent-guaranteed subsidiary on segments when they are in industries whose cash flows are less correlated with the capital expenditures of industries of other segments within the firm. Specifically, I run the following multinomial logit regression on my sample of segments within diversified firms (the same 9 Such covenants most commonly restrict dividends to be to 50% of net income less 100% of cumulative losses. In some cases, covenants completely prohibit dividends. In some cases dividend restrictions are less strict. My results are not sensitive to alternate definitions of the dividend restrictions variable.

341

sample as used in Section 4.1): Pðwd ¼ 1; 2Þ ¼ Lða þ b1 mkttobk_difference þ b2 stddevðmkttobkÞ þ b3 stddevðcf _volÞ þ b4 cf _corr þ GControls1 Þ þ  ðM-Model 7Þ

where all variables, except cf_capx_corr, are the same as in Section 4.1. I define cf_capx_corr as the correlation of a segment’s industry cash flows with the industry capital expenditures of all other segments within the firm. When there are more than two segments within a firm, I take the asset-weighted average of each correlation. I define industry cash flow as the asset-weighted average ratio of EBITDA/Assets for all standalone firms in a three-digit NAICS industry. I define industry capital expenditures as the asset-weighted average of the ratio of capital expenditures to assets of all standalone firms in a threedigit industry. I compute the correlations between industry pairs using five years of lagged quarterly data. If endogeneity is driving my results in Section 4.2, I expect the coefficient on cf_capx_corr in M-Model 7 to be negative. Contrary to this prediction, in untabulated results, I find the coefficient is not statistically different from zero. It is also possible that firms do not divert capital from segments with guaranteed debt because it is efficient not to do so, as they have the highest mkttobk in the firm. This alternative hypothesis, however, runs counter to Shin and Stulz (1998), who find that diversified firms on average do not protect the capital budgets of high mkttobk segments. To give further assurance that endogeneity is not driving my results, I examine whether having a debt in its own capital structure affects the sensitivity of an individual segment’s investment to other segment cash flows. In untabulated results, I fail to find an effect. Taken together with my results in Section 4.2.1, this implies that subsidiary debt is associated with fewer cash flows being diverted out of an indebted segment. However, subsidiary debt does not reduce the extent to which other nonindebted segment cash flows are diverted into the indebted segment. 5. Discussion of alternative rationales for subsidiary debt Federal taxes. Subsidiary debt has no advantage relative to parent debt in reducing federal tax liability as long as the parent files a consolidated tax return or the subsidiary is a pass-through entity. A corporate subsidiary at least 80% owned and domestically incorporated is eligible for tax consolidation.10 A subsidiary organized as a partnership, limited liability company (LLC), or unlimited liability entity that passes through income to its owners pays no corporate income tax.11 Using public filings, I find that all but four subsidiaries in my sample are either passthrough entities or eligible for tax consolidation. State taxes. Firms operating in multiple states can reduce their state tax liability by placing debt on 10 11

Scholes, Wolfson, Erickson, Maydew, and Shevlin, (2002, p. 297). Ibid, pp. 64 and 80.

ARTICLE IN PRESS 342

A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

subsidiaries in high tax states. However, intra-company debt suffices for this purpose (Brunori, 2001, p. 113; Pomp, 1998). Lower risk. Subsidiary debt at the safer division has a claim on that division senior to that of all other securities. Thus, such debt is sometimes safer than parent debt and carries a lower required rate of return. Use of such debt, however, does not lower a firm’s total cost of capital. At best, it transfers risk from subsidiary creditors to holders of other securities, leaving total risk and cost of capital unchanged.

6. Conclusion This study empirically examines why firms sometimes separately incorporate some of their non-financial operating divisions as subsidiaries and include debt in these subsidiaries’ capital structures. I find that firms tend to include nonguaranteed subsidiary debt in their divisions’ capital structures when their divisions vary more in operating risk. This result supports the Kahn and Winton (2004) model, in which firms use nonguaranteed subsidiary debt to reduce asset substitution incentives. I also find some evidence that cash-cow divisions, those with high cash flows but fewer relative growth options, also tend to have nonguaranteed subsidiary debt outstanding. Though it is not as robust to alternative specifications of growth options, this result suggests firms use nonguaranteed subsidiary debt as a means to control the wastage of free cash flows in their cash cows without inducing underinvestment in their growth divisions. I also find that divisions within a firm tend to have parent-guaranteed subsidiary debt in their capital structures when they have investment opportunities better than the rest of the firm. Furthermore, I find that cash flows of divisions with parent-guaranteed debt have a smaller effect on investment in the rest of the firm than do the cash flows of divisions without such debt. Thus, parent-guaranteed debt appears to reduce the extent to which a division’s cash flows are diverted and invested in other divisions. Taken together, this set of results provides evidence that firms use guaranteed subsidiary debt, at least in part, to protect their investment opportunity-rich subsidiaries from the underfunding and poaching problems modeled in Scharfstein and Stein (2000) and Rajan, Servaes, and Zingales (2000). Specifically how parent-guaranteed subsidiary debt protects firms’ growth divisions from poaching is not clear. I fail to find evidence that debt covenants restricting dividends to the parent have an effect, though my test has low power as it is limited to public debt. However, there are other potential mechanisms. For example, by issuing public debt backed by a high-growth division, a firm commits to publicly disclosing that division’s cash flows and capital expenditures to outsiders, inviting scrutiny if the division gets poached. As the creditors of the subsidiary are senior to parent creditors, they have an interest in ensuring that the subsidiary’s investment budget is not underfunded. Finally, it is possible that by allowing a division to raise its own capital, and guaran-

teeing its debt, a firm is signaling its desire to ensure that the division is fully funded and allowed to take advantage of all its profitable investment opportunities. Thus, it is likely that financing a subsidiary with its own parentguaranteed debt is part of a package of measures firms use to nurture and protect their growth divisions. Exploring what other measures lie in this package will prove to be an exciting and fruitful area of future research in corporate finance. References Barclay, M., Smith, C., 1995a. The maturity structure of corporate debt. Journal of Finance 50, 609–631. Barclay, M., Smith, C., 1995b. The priority structure of corporate liabilities. Journal of Finance 50, 899–917. Berger, P., Ofek, E., 1995. Diversification’s effect on firm value. Journal of Financial Economics 37, 39–65. Billet, M., King, T., Mauer, D., 2007. Growth options and the choice of leverage, debt maturity and covenants. Journal of Finance 62, 697–792. Brunori, D., 2001. State Tax Policy: A Political Perspective. The Urban Institute Press, Washington. Campa, J., Kedia, S., 2002. Explaining the diversification discount. Journal of Finance 57, 1731–1762. Census Bureau, 1993. Economic classification policy committee issues. Conceptual Issues, paper 1. Available at /www.census.gov/epcd/ www/naicsdoc.htmS. Chava, S., Roberts, M., 2008. How does financing impact investment? The role of debt covenant violations. Journal of Finance 63, 2085–2121. Chemmanur, T., John, K., 1996. Optimal incorporation, structure of debt contracts, and limited-recourse project financing. Journal of Financial Intermediation 5, 372–408. Chevalier, J., 2004. What do we know about cross-subsidization? Evidence from the investment policies of merging firms. Advances in Economic Analysis & Policy 4, 1–27. Childes, P., Mauer, D., Ott, S., 2005. Interactions of corporate financing and investment decisions: the effects of agency conflicts. Journal of Financial Economics 76, 667–690. Flannery, M., Houston, J., Venkataraman, S., 1993. Financing multiple investment projects. Financial Management 22, 161–172. Gomes, J., Livdan, D., 2004. Optimal diversification: reconciling theory and evidence. Journal of Finance 59, 507–535. Graham, J., Lemmon, M., Wolf, J., 2002. Does corporate diversification destroy value? Journal of Finance 57, 695–719. Guedes, J., Opler, T., 1996. The determinants of the maturity of corporate debt issues. Journal of Finance 51, 1809–1833. Henderson, B., Jegadeesh, N., Weisbach, M., 2006. World markets for raising new capital. Journal of Financial Economics 82, 63–101. Jensen, M., 1986. Agency costs of free cash flow, corporate finance, and takeovers. American Economic Review 76, 323–329. Jensen, M., Meckling, W., 1976. Theory of the firm: managerial behavior, agency, and ownership structure. Journal of Financial Economics 3, 305–360. John, T., 1993. The optimality of spin-offs and allocation of debt. Journal of Quantitative and Financial Analysis 28, 139–160. John, T., John, K., 1991. Optimality of project financing: theory and empirical implications in finance and accounting. Review of Quantitative Finance and Accounting 1, 51–74. Johnson, S., 2003. Debt maturity and the effects of growth opportunities and liquidity risk on leverage. Review of Financial Studies 16, 209–236. Kahn, C., Winton, A., 2004. Moral hazard and optimal subsidiary structure for financial institutions. Journal of Finance 59, 2537–2575. Kensinger, J., Martin, J., 1988. Project finance: raising money the oldfashioned way. Journal of Applied Corporate Finance 1, 69–81. Lamont, O., 1997. Cash flow and investment: evidence from internal capital markets. Journal of Finance 52, 83–109. Lamont, O., Polk, C., 2002. Does diversification destroy value? Evidence from industry shocks. Journal of Financial Economics 63, 51–77. Leland, H.E., 2007. Financial synergies and the optimal scope of the firm: implications for mergers, spinoffs, and structured finance. Journal of Finance 62, 765–807. Lang, L., Stulz, R., 1994. Tobin’s q, corporate diversification, and firm performance. Journal of Political Economy 102, 1248–1280.

ARTICLE IN PRESS A.C. Kolasinski / Journal of Financial Economics 94 (2009) 327–343

Maksimovic, V., Phillips, G., 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. Mian, S., Smith, C., 1992. Accounts receivable management policy: theory and evidence. Journal of Finance 47, 169–200. Morellec, E., 2001. Asset liquidity, capital structure, and secured debt. Journal of Financial Economics 61, 173–206. Myers, S., 1977. Determinants of corporate borrowing. Journal of Financial Economics 5, 147–175. Nanda, V., 1991. On the good news of equity carve-outs. Journal of Finance 46, 1717–1737. Noe, T., 1998. Creditor rights and multinational capital structure. Unpublished working paper, Tulane University. Office of Management and Budget, 1994. Standard industrial classification replacement, economic classification policy committee. Available at /http://www.census.gov/epcd/naics/ascii/naicsfr2.txtS. Ozbas, O., Selvili, Z., 2006. Integration and corporate investment: empirical evidence on rigid capital budgets. Unpublished Working Paper, USC. Pomp, R., 1998. The future of the state corporate income tax: reflections (and confessions) of a tax lawyer. In: David, B. (Ed.), The Future of State Taxation. Urban Institute Press, Washington. Rajan, R., Servaes, H., Zingales, L., 2000. The cost of diversity: the diversification discount and inefficient investment. Journal of Finance 55, 35–80. Rogers, W., 1993. Regression standard errors in clustered samples. Stata Technical Bulletin 13, 19–23. Samson, S., Sprinzen, S., Dubois-Pelerin, E., Pfeil, K., 2005. Corporate criteria—parent/subsidiary links, in Standard & Poor’s corporate ratings criteria. Available at /http://www2.standardandpoors.com/ spf/pdf/fixedincome/CorporateRatings2005.pdfS. Scharfstein, D., 1998. The dark side of internal capital markets II: evidence from diversified conglomerates. NBER Working Paper # 6532. Scharfstein, D., Stein, J., 2000. The dark side of internal capital markets: divisional rent-seeking and inefficient investment. Journal of Finance 55, 2537–2564. Schoar, A., 2002. The effects of corporate diversification on productivity. Journal of Finance 57, 2379–2403.

343

Scholes, M., Wolfson, M., Erickson, M., Maydew, E., Shevlin, T., 2002. Taxes and Business Strategy: A Planning Approach, second ed. Prentice-Hall, Upper Saddle River, NJ. Shin, H., Stulz, R., 1998. Are internal capital markets efficient? Quarterly Journal of Economics 113, 531–552. Slovin, M., Sushka, M., 1997. The implications of the equity issuance decision within a parent-subsidiary structure. Journal of Finance 52, 841–857. Smith, C., Watts, R., 1992. The investment opportunity set and corporate financing, dividend, and compensation policies. Journal of Financial Economics 32, 263–292. Smith, C., Warner, J., 1979. On financial contracting: an analysis of bond covenants. Journal of Financial Economics 7, 117–161. Stulz, R., Johnson, H., 1985. An analysis of secured debt. Journal of Financial Economics 14, 501–521. Stumpp, P., Rottino, W., Fanger, D., Stephan, R., Carter, D., 2003. Rating nonguaranteed subsidiaries: credit considerations in assigning subsidiary ratings in the absence of legally binding parent support, in Moody’s Rating Methodology. Credit Policy & Methodologies Directory at /www.moodys.comS. Thomson, R., 1991. Piercing the corporate veil: an empirical study. Cornell Law Review 76, 1036–1065. Villalonga, B., 2004. Diversification discount or premium? New evidence from the business information tracking series. Journal of Finance 59, 479–506. Vijh, A., 2002. The positive announcement returns of equity carve-outs. Journal of Business 75, 153–190. Vijh, A., 2006. Does a parent-subsidiary structure enhance financing flexibility? Journal of Finance 61, 1337–1360. Williams, R., 2000. A note on robust variance estimation for clustercorrelated data. Biometrics 56, 645–646. White, W., 1974. Debt management and the form of business financing. Journal of Finance 29, 1467–1484. Whited, T., 2001. Is it inefficient investment that causes the diversification discount? Journal of Finance 56, 1667–1691. Wooldridge, J., 2002. Econometric Analysis of Cross Section and Panel Data. MIT Press, Cambridge, MA. Wulf, J., 2009. Influence and inefficiency in the internal capital market: theory and evidence., Journal of Economic Behavior and Organization, forthcoming.