Special Purpose Entities: Empirical Evidence on ... - SSRN papers

1 downloads 0 Views 357KB Size Report
E-mail: sanjay.gupta@asu.edu. This version: January 2006. Abstract. This paper investigates the usage, determinants and earnings effect of special purpose ...
Special Purpose Entities: Empirical Evidence on Determinants and Earnings Management* Mei Feng Katz Graduate School of Business, University of Pittsburgh E-mail: [email protected] Jeffrey D. Gramlich University of Southern Maine School of Business and Copenhagen Business School E-mail: [email protected] Sanjay Gupta W. P. Carey School of Business, Arizona State University E-mail: [email protected]

This version: January 2006

Abstract This paper investigates the usage, determinants and earnings effect of special purpose entities based on a large sample of firms from 1994 to 2004. We find that SPE use has grown substantially in recent years and that the number of SPEs is related to economic motivations, financial reporting objectives and corporate governance. Specifically, SPE use is significantly higher for firms with greater need for funds and higher marginal tax rate. We also find that SPE use increases with closeness to debt covenants, CEO’s bonus and weak corporate governance. Finally, we document that the evidence is consistent with SPEs set up for financial reporting purposes playing a role in managing accruals and earnings upwards, whereas the same does not appear to be the case for SPEs set up mainly for economic reasons. *

The authors gratefully acknowledge John Graham, Jim Largay, Andy Leone, Cathy Shakespeare, and The Corporate Library for sharing various data used in this study. Helpful comments from Anwer Ahmed, Patty Dechow, Harry Evans, Liming Guan, Emre Karaoglu, Mary Lea McAnally, Sarah McVay, Lil Mills, Kaye Newberry, Tom Omer, John Robinson, Cathy Shakespeare, Marjorie Shelley, Anup Srivastava, Terry Shevlin, Tom Stober, Jake Thomas and Dan Weimer, and workshop participants at the University of Pittsburgh, Texas A&M University, University of Connecticut and the 2005 American Accounting Association annual meeting are gratefully appreciated.

Special Purpose Entities: Empirical Evidence on Determinants and Earnings Management 1.

Introduction In this study we examine firms’ motivations for using off-balance sheet entities known as

special purpose entities, special purpose vehicles, or variable interest entities (collectively referred to as “SPEs”). Specifically, we investigate the extent that SPE sponsorship is explained by economic arguments, financial reporting motivations, and corporate governance qualities. We then examine whether SPEs formed primarily for financial reporting reasons are associated with discretionary accruals and unexpected return on equity. We base our analysis on a proxy of SPE use that can be applied to a broad cross-section of firms over time. An SPE is a legally distinct entity with a limited life, created to carry out a narrowly defined pre-specified activity, typically to benefit only one “sponsor” company (Coallier [2002], Dharan [2002], Hargraves and Benston [2002], FASB [2002, Statement 125 para. 26]). Isolating the cash flows and business risks to the pre-specified purpose allows the sponsor to market debt and equity in the SPE as investments that meet specific clientele needs, thereby providing the sponsor with increased liquidity and lower capital costs. In addition, an SPE is attractive from a financial reporting standpoint because the the SPE is excluded from the sponsor’s balance sheet as long as outside investors contribute “substantive residual equity investment,” which until recently was defined as at least three percent of the SPE’s total capitalization.1 More importantly, the sponsor’s income statement could also be managed.

1

This guidance can be traced to the FASB’s Emerging Issues Task Force (EITF) bulletin No. 90-15 (FASB [1990]), which provided that in leasing transactions three percent was the minimum investment required by outside owners other than the lessee. The SEC staff’s response to the EITF 90-15 effectively extended the three percent test to other transactions. We discuss the accounting guidelines in detail later.

1

Although SPEs can serve legitimate business purposes, policy makers have been concerned with their use for financing arrangements that are kept off-balance sheet because of the consequences of masking firms’ financial risks from investors and creditors. Enron’s sudden and spectacular demise, due in large part to its aggressive use of SPEs to opportunistically manage its debt and earnings (Powers [2002]), vividly illustrates the potential consequences. Financial commentators (e.g., Bryan-Low and Brown [2002], Covert [2002], Cowan and Talley [2002]), including Enron’s auditor (Berardino [2001]), caution investors that other companies are likely to be using SPEs to achieve Enron-style opportunistic objectives. Regulators’ recent responses are recognition of the importance of these concerns. The SEC, acting under the mandate of section 401(a) of the Sarbanes-Oxley Act of 2002, now requires disclosure of offbalance sheet entities, and the FASB has increased the minimum outside investment requirement from three to 10 percent and added substantial quantifiable and qualitative “at risk” requirements to avoid consolidating the SPE.2 We measure SPE use as a count of the limited partnerships, limited liability companies or trusts included in the list of subsidiaries and affiliates in Exhibit 21 of the SEC Form 10K. This proxy applies to a broad cross-section of firms over a time period when few disclosure requirements concerning SPEs existed. We choose this proxy because SPEs are typically organized using one of these legal structures (Dharan [2002], Clauss and Reed [2003], Soroosh and Ciesielski [2004]), which was corroborated by the staff of the SEC’s Office of the Chief Accountant and anecdotal evidence surrounding Enron (Powers [2002], Goldin [2003]) and Adelphia (Securities and Exchange Commission v. Adelphia Communications et al. [2002]). In particular, Enron’s SPE-type entities listed in its Exhibit 21 were mostly organized as limited 2

These new requirements are generally effective for firms’ 2003 or 2004 fiscal years and are contained in FASB Interpretation No. 46 (R).

2

partnerships and grew dramatically from 31 in 1994 to 850 in 2000 (see Appendix A). Many of these partnerships played a key role in hiding losses and concealing debt, which when disclosed led to Enron’s collapse.3 Because not all affiliated LPs, LLCs and trusts likely meet FASB’s definition for either “special purpose entity” or “variable interest entity,” we perform several validity checks to evaluate whether our proxy reasonably captures SPE activity. Using a sample of 6,892 unique firms from 1994 through 2004, we obtain 25,844 firmyear observations for which Exhibit 21 and certain other data are available. During this 11-year period, we find a six-fold monotonic increase in the percent of observations reporting at least one SPE-type entity (from under 10 percent in 1994 to nearly 60 percent in 2004). Further, the mean (median) number of SPEs reported among firm-years with at least one SPE grew monotonically from 4.56 (1) in 1996 to 16.05 (4) in 2004. The percentage of SPE use is increasing in firm size, consistent with larger firms being better able to deal with the complexity of structured financing arrangements (SEC [2005]). In addition, SPE use is highest among industry groups that tend to be leasing-activity intensive, such as trading, real estate and construction. We also find a relatively high use of SPEs in banking and telecommunications, which is consistent with these industries providing new areas for applying SPEs during the 1990s, such as the securitization of financial assets and broadband capacity (Dharan [2002]). Our Tobit regression results for the determinants of SPE use show that, controlling for firm size and industry membership, the number of SPEs is related to economic motivations, financial reporting objectives and corporate governance. Specifically, SPE use is increasing in

3

Powers [2002] details Enron’s SPE structures and transactions and, especially, the role of SPEs such as Raptor, Raven, and Whitewing, organized as limited partnerships and listed in Exhibit 21 of the company’s 2000 10K, in concealing debt and avoiding losses. See also, Dharan [2002], Emshwiller and Smith [2002] and Yale [2002] for a discussion of Enron’s use of SPEs.

3

firms’ need for funds and marginal tax rate. SPE use also increases with closeness to debt covenants and the CEO’s bonus, but strong corporate governance mitigates SPE use. To shed light on whether SPEs formed for different objectives are used to manage earnings upwards, we first parse the number of SPEs for a firm year into those predicted by the economic and financial reporting motivations using the Tobit model results. We then investigate the relation between these predicted components and firms’ discretionary accruals and unexpected earnings. Our overall evidence is consistent with SPEs set up for financial reporting purposes playing a significant role in managing accruals and earnings upwards, whereas the same does not appear to be the case for SPEs set up mainly for economic reasons. Our study builds on the limited extant research concerning SPEs. Prior studies tend to conduct pointed examinations of SPE use in particular transactions, such as research and development financing (e.g., Shevlin [1987], Beatty, Berger and Magliolo [1995]) and asset securitizations (e.g., Dechow, Myers and Shakespeare [2005], Karaoglu [2005]). Although the transaction-level focus has the merit of pinpointing the circumstances in which SPEs are used, it also generally necessitates relying on small samples or limited time periods. Both of these research design issues limit generalizability, which has emerged as an important concern in the effort to determine whether high-profile corporate scandals (e.g., Enron, Adelphia, etc.) are isolated examples or whether the opportunistic use of SPEs is more pervasive (e.g., Soroosh and Ciesielski [2004]; SEC [2005]). We avoid a transaction-level focus and conduct a broad examination of the SPE phenomenon. In this respect, our study is nearest to Mills and Newberry [2005], who use the book-tax difference in interest expense to proxy off-balance sheet financing; they obtain a larger sample than earlier studies that focus on a specific type of transaction. However, the book-tax

4

difference proxy is best viewed as an indirect test of the SPE phenomenon, it excludes SPEs for which the book-tax treatment conforms and requires access to confidential tax return data. In contrast, our SPE proxy is not constrained by book-tax conformity issues, is constructed from publicly-available data, and we believe identifies SPEs more directly. Perhaps more significantly, Mills and Newberry’s [2005] focus is on the role of firms’ bond ratings and leverage ratios on off-balance sheet financing, but they do not consider the effects of corporate governance nor do they examine earnings management. In this regard, our study is similar to Dechow et al. [2005] who examine corporate governance and earnings management in the context of asset securitizations, but only for a limited sample of 80 firms with post-September 2000 data when the new securitization disclosure requirements became effective. The remainder of the paper is organized as follows. Section 2 provides the institutional background surrounding SPEs and their consequences on earnings. Section 3 defines our SPE proxy, explains our sampling procedures, and describes the sample. Section 4 discusses the empirical model and defines the variables. Section 5 reports the regression results of the determinants of SPE use, and section 6 describes the results of the earnings management tests. Section 7 concludes with a summary of our findings.

2.

Background

2.1.

FORMATION AND STRUCTURE OF AN SPE SPEs have long been used in a variety of arrangements that have off-balance sheet

implications, although their use in leasing and securitization transactions is widespread and most well-known (Hodge [1998], SEC [2005]). Figure 1 outlines a typical SPE formation with the steps occurring in coordinated fashion rather than in chronological order, and certain details of the steps differing depending on the SPE’s function (i.e., leasing or asset securitization). 5

Typically, the sponsor forms a separate legal entity, and unrelated investors provide capital in exchange for residual equity interests in the SPE. The SPE receives a loan that is used to purchase an asset or class of assets, often in a transaction with the sponsor, and the asset(s) are collateral. The sponsor typically reduces the risk of outside SPE investors, both creditors and residual equity holders, by guaranteeing the loan or agreeing to incur a portion of the potential losses associated with the assets transferred to the SPE, or both. 2.2.

NATURE OF AND OBJECTIVES FOR FORMING AN SPE SPEs are unique in the sense that their special purpose is defined by a charter that

specifically limits the scope of the entity’s activities to just those directly related to that purpose. According to the FASB, the term “SPE” is so narrow in scope that it does not apply to operating businesses with employees and equity sufficient to operate without support from a business entity other than its owners (FASB [2002, para. 7]). Most SPEs have all of their significant activities “pre-programmed,” such that they are commonly referred to as “brain-dead” (Dharan [2002]). An SPE is also usually “bankruptcy-remote” in that the lender is insulated from the insolvency or bankruptcy risks of the sponsor, with the lower credit risk resulting in lower borrowing costs. In addition, SPEs normally provide tax efficiency by passing through tax deductions (income items) to the investors who are in the best positions to utilize (report) them. All of these unique characteristics enable SPEs to serve legitimate business purposes. We identify these objectives as the economic arguments for forming SPEs and include within them aspects of SPEs that can increase the sponsor firm’s value. The unique characteristics of SPEs, coupled with the accounting rules for consolidated financial statements, provide another attraction for SPEs – sponsors can exclude these entities from their financial statements. Off-balance sheet treatment is possible so long as the SPE meets

6

certain conditions. The principal condition has been based on the level of third-party investment in the equity of the SPE. Until 2003, U.S. GAAP employed the unilateral control approach based primarily on voting rights to determine whether assets and liabilities would be consolidated (FASB [2003]). The nature of SPEs, however, renders voting control irrelevant because SPEs’ activities are circumscribed by their charters and contracts are built into their operating structure upon formation (i.e, they are “pre-programmed”). As a result, a sponsor essentially controls an SPE’s actions despite not having voting control. Hence, FASB has recently applied a more stringent “risks and rewards” test under which consolidation of SPEs is avoided only if sponsors transfer sufficient risks and rewards to independent third parties (FASB [2003]). Until FASB issued Interpretation No. 46, the sufficiency test was met if the third party’s residual equity investment at risk was at least three percent of the SPE’s total capitalization. In response to the perceived abuses involving SPEs, FASB issued new principles-based guidance for consolidation of SPEs that emphasizes economic substance over form.4 Circumventing the consolidation rule allows the sponsor to hide within an SPE debt for which the sponsor is ultimately liable, which can be substantial especially because SPEs are typically thinly capitalized.5 Moreover, an SPE also enables its sponsor to manage accruals and cash flows upward through related-party transactions with SPEs.6 Thus, apart from the economic 4

The new guidance is contained in FIN 46R issued in December 2003 that requires consolidation based on the variable interest model that uses risks and rewards rather than voting control. Under this new model, a “variable interest entity” (VIE) must be consolidated in the financial statements of the “primary beneficiary”. It also increases the minimum third party investment from three to 10 percent. 5

FASB also issued FIN 45 (effective December 15, 2002) to require loan guarantors to recognize the fair market value of loan guarantees as a liability on the balance sheet, replacing previous guidance that only required liability recognition if payment was both probable and estimable.

6

Enron’s court-appointed bankruptcy examiner found that Enron was able to keep about $14 billion of debt off its balance sheet through structured finance transactions involving SPEs (Batson [2003]). Similarly, a Special Investigative Committee on Enron reported that transactions with certain SPEs “allowed Enron to conceal from the market very large losses resulting from Enron’s merchant investments by creating an appearance that those investments were hedged” (Powers [2002, p. 4]). This report also illustrated Enron’s use of derivatives together with SPEs to disguise debt financing as operating cash flows.

7

uses of SPEs, financial reporting concerns apparently are an important additional objective for forming SPEs. We refer to financial reporting determinants as those objectives for forming SPEs that motivate management to intervene in the financial reporting process. As the SEC study of SPEs [2005] notes, the corporate scandals involving SPEs exposed weaknesses not only in financial reporting but also in corporate governance. Dharan [2002] argues that organizational commitment to earnings management is required to use financial engineering techniques such as SPEs to manage financial statements. Thus, we believe that strong corporate governance characteristics should mitigate the propensity of firms for establishing SPEs to meet financial reporting objectives. This reasoning also suggests a potential interactive effect of financial reporting variables with corporate governance, and so we incorporate that in the empirical analysis. Based on the above discussion, our conceptual model of SPE use is as follows: SPE use = f (economic variables, financial reporting variables, corporate governance variables, and other controls). The economic variables proxy the need for funds, tax benefits and risk clienteles; the financial reporting variables capture contracting arguments and external financing needs; corporate governance variables reflect board composition; and controls include firm size, and industry membership. We develop empirical proxies for these concepts in section 4. 2.3.

THE IMPACT OF SPES ON ACCRUALS AND AND CASH FLOWS Traditional earnings management focuses exclusively on accruals. SPEs, however, allow

managers to manage earnings through cash flows as well. For example, Dechow et al. [2005] show that in asset securitization transactions structured through SPEs, sponsors receive cash upon consummation and the amount of the proceeds depends upon the relative degree of security

8

that the sponsor provides the SPE investors. Further, managers seeking to increase earnings have an incentive to overvalue the sponsor’s retained interest in the securitized assets, thereby increasing the sponsor’s gain. Similarly, in a sale-leaseback transaction a sponsor sells appreciated property to an SPE for cash. The SPE may simultaneously pledge the property as collateral for a loan guaranteed by the sponsor. By defining the extent of the loan guarantee, the sponsor exercises influence in determining the value of the loan proceeds. Further, through control over the deemed fair market value of the transaction, the sponsor also manages the amount of the gain recognized on the transaction.7 After selling the property to the SPE, the sponsor typically leases the property back from the SPE, enabling the sponsor to continue using the property in its operations. Although we expect SPEs created for financial reporting reasons to increase earnings from accruals, focusing only on accruals may cause us to miss some “management intervention in the external financial reporting process,” as Schipper [1989] defines earnings management, since SPEs can positively impact cash flows. Hence, we examine the impact of SPEs on both accruals and return on equity (ROE), a measure that includes both accruals and cash flows.

3.

Measure of SPE Use, Sample Selection, and Descriptive Statistics

3.1.

MEASURE OF SPE USE – DEPENDENT VARIABLE Our measure of SPE use is the count of the number of subsidiaries or affiliates listed in

Exhibit 21 of SEC Form 10K filings with names that contain “Limited Partnership,” “L.P.,”

7

SFAS No. 28 (FASB [1979]) limits the amount of gains, but not losses, recognized in all sale-leaseback transactions. When the present value of leaseback payments exceeds 10 percent of the asset’s fair market value, SFAS 28 stipulates that only a portion of the realized gain can be recognized immediately, with the amount deferred depending upon whether the present value of the lease payments exceeds 90 percent of the asset’s fair market value and whether the lease is treated by the lessee as an operating or capital lease.

9

“LP,” “LLC,” “L.L.C.,” or “trust.”8 SEC Regulation S-K Subpart 229.601(b)(21)(i) requires firms to list “subsidiaries” in the annual 10K filing. For this purpose, SEC Regulation S-X Part 210.1-02(x) defines “subsidiary” as an affiliate that the firm controls directly or indirectly. Regulation S-X Part 210.1-02(b) defines an “affiliate” as a “person” who directly, or indirectly through intermediaries, controls, is controlled by, or is under common control with the person specified. A “person” for this purpose can be an individual or any type of business or not-forprofit entity [Regulation S-X Part 210.1-02(q)]. The SEC defines “control” as “the possession, direct or indirect, of the power to direct or cause the direction of the management and policies of a person, whether through the ownership of voting shares, by contract, or otherwise” [Regulation S-X Part 210.1-02(g), emphasis added]. Because “control” can occur via contract, special purpose entities, as defined earlier, fall within the SEC’s definition of a subsidiary. Consistent with these requirements, Enron listed in its Exhibit 21 many of the SPEs that have been identified as off its balance sheet and used to manage debt and earnings (see Figure 2). To evaluate our SPE activity measure’s ability to capture the phenomenon of off-balance sheet SPEs, we perform several validity checks and report these results later. 3.2

SAMPLE SELECTION. We begin with the 33,244 firm years that list subsidiaries and affiliates in Exhibit 21 to

the 10-K filings in the SEC’s EDGAR database that were filed between December 1994 and December 17, 2004. After deleting observations without fiscal year end and SIC data, 25,844 observations for 6,892 firms remain.9 We use this sample to provide large-sample descriptive

8

In practice some firms list their subsidiaries and affiliates in Exhibit 22 of Form 10K instead of Exhibit 21. In that case, we use the subsidiaries and affiliates listed in Exhibit 22 instead of Exhibit 21.

9

EDGAR files only report companies’ names and CIKs. To identify which of these firms are on Compustat, we match the Edgar CIKs with its counterpart company identifier on Compustat, GVKEY. We thank Andy Leone for helping us make this match.

10

evidence on the prevalence and extent of SPE use. Because our regression tests have greater data requirements, we delete observations with insufficient financial statement, executive compensation, daily stock return, marginal tax rate, or corporate governance data. This reduces the number of observations to 12,632, 4,563 and 2,403 firm-years, respectively, when we require data to calculate the economic, financial reporting and corporate governance variables. Panel A of Table 1 summarizes the sample selection. We obtain financial reporting, executive compensation and stock return data from Compustat, Execucomp and CRSP, and corporate governance data from the Investor Responsibility Research Center, Inc. (IRRC) for the years 1996 to 2003. Though we utilize only a few data items, IRRC data summarizes public information sources into a wide array of variables concerning governance provisions for about 2,000 public U.S. corporations. 3.3.

VALIDITY CHECKS FOR SPE MEASURE. We conduct three checks to validate our SPE activity measure. First, we search for the

terms “special purpose entity” and/or “variable interest entity” in the entire Form 10Ks filed electronically with the SEC between November 29, 1993, and December 17, 2004. As Panel B of Table 1 shows, approximately 13.7 percent of the filings (4,539 of 33,046 firm-years) mention either term at least once in Form 10K. Of these 4,539 firm-years, 63.3 percent (2,874 of 4,539) of the Exhibit 21s list at least one LP, LLC, or trust. his percentage falls to 36.8% (10,501 of 28,507) for firms that do not mention either of the two terms in their 10K. In other words, firms stating either “special purpose entity(ies)” or “variable interest entity(ies)” in the text of their Form 10K are nearly twice as likely to list a LP, LLC, or trust in Exhibit 21.10 10

We repeat this validity check on 10Ks filed after the Enron bankruptcy on December 2, 2001. It is likely that more firms started mentioning SPEs or VIEs to specifically state that they are not using these types of off-balance sheet entities. While the percentage of firms mentioning SPEs/VIEs goes up nearly threefold, about the same percentage (63.6 percent) of these firms list SPE-type entities in their Exhibit 21.

11

Second, because SPEs are often used in connection with asset securitizations, we examine the frequency with which our SPE measure occurs in a sample of 66 hand-collected asset securitization transactions for at least $200 million that occurred during 2001 and 2002.11 Of the 53 observations for which we find a match in our sample for those years, 46 (86.9 percent) list at least one limited partnership, limited liability company, or trust in their Form 10K Exhibit 21. This percentage is nearly double the 46 percent for our sample in 2001 and 2002 that reports at least one SPE-type entity in Exhibit 21, suggesting that our SPE measure likely captures the type of activity typically related to off-balance sheet entities used for asset securitization purposes. Our final validity check is based on Dechow et al.’s [2005] evidence that firms with stronger corporate governance are less likely to recognize gains from asset securitizations, consistent with managers of these firms making more conservative financial reporting assumptions. Since SPEs facilitate favorable accounting reflections of asset securitizations, as well as leasing arrangements, Dechow et al.’s finding suggests that corporate governance ratings are likely negatively related to the number of SPEs that firms sponsor. We use governance ratings provided by The Corporate Library for the year 2001, translated from a five-point A-E grading scale to a 5-1 numerical scale (i.e., A=5 and E=1). For the 733 observations in our 2001 sample for which we could obtain the governance ratings, the Pearson (Spearman) correlation between this rating and our SPE measure is -0.225 (-0.305) with a p-value of .03 ( 0 ⎪ SPEi ,t ⎨ ⎪ ⎪= 0 if X i ,t −1 ' β + Yi ,t −1 ' γ + Z i ,t −1 ' λ + Vi ,t −1 ' κ + ε i ,t ≤ 0 ⎩

(1)

where: X i ,t −1 ' β = α + β1 FUNDSi ,t −1 + β 2 CLTDi ,t −1 + β 3 MTRi ,t −1 + β 4 RISK i ,t −1

Yi ,t −1 ' γ = γ 1 LEVi ,t −1 + γ 2 INTCOVi ,t −1 + γ 3 BONUSi ,t −1 + γ 4 DEBTISSi ,t +1 + γ 5 STOCKISSi ,t +1 Z i ,t −1 ' λ = λ1 DIRINDi ,t −1 + λ2 INDSH i ,t −1 + λ3 BUSYi ,t −1 Vi ,t −1 ' κ = κ1 SIZEi ,t −1 + κ 2 INDU _ PERCi ,t −1 + κ 3YEARi ,t . 12

The linear regression model yields biased and inconsistent parameter estimates for censored data because it fails to account for the qualitative difference between the limit (zero) and non-limit (continuous) observations. Maximum likelihood estimation overcomes this bias by accommodating a mixture of both densities (for the non-limit observations) and cumulative densities (for the limit observations). See generally Greene [1993, p.694-697] and Kennedy [1998, p.249-256].

14

We discuss the motivation, definition and predicted sign of each variable in the next subsection and provide a list of variable definitions in Table 4. With respect to the interactions between governance and financial reporting examined in model 4, the fact that we utilize three governance variables complicates matters. To keep the specification tractable, we interact only one of the corporate governance variables at a time with each of the financial reporting variables and discuss this more fully later (see section 5.2). 4.1.

ECONOMIC DETERMINANTS Need for funds. As discussed before, SPEs can help sponsors to raise capital. SPEs’

charters specify in advance the cash flows and risks of a business venture and isolate the SPE assets from the general corporate purpose of the sponsor. As a result, an SPE can raise funds at lower costs. On the other hand, as Wolfson [1985] points out and Enron demonstrates, SPEs can be a costly means of raising capital because of the high contracting costs that result from management’s conflicts of interest. Overall, we predict that firms will employ SPEs when the need for additional capital is greatest. FUNDS and CLTD are the two financing needs variables.13 FUNDS is the supply of internally generated funds, defined as the sum of cash flow from operations and investing activities divided by average total assets. Following Beatty et al. [1995], we expect a negative coefficient on FUNDS because firms with greater internally generated funds are less likely to raise capital via SPEs. CLTD, measured as long term debt due within one year divided by total

13

Diminished debt capacity, as measured by increased financial leverage, is also likely to be positively related to the demand for SPEs as a capital source. Based on earlier research, however, we employ leverage as a measure of tightness to debt covenants, and recognize the limitations of interpreting the outcome as being due unilaterally to either decreased debt capacity or increased tightness of financial covenants.

15

assets, serves as a proxy for renegotiation costs. Consistent with Beatty et al., we predict a positive coefficient on CLTD. Tax benefits. Another reason for forming an SPE is the ability to allocate tax benefits (costs) to taxpayers who can most efficiently utilize (report) them. In the LP, LLC and real estate mortgage investment conduit (REMIC) contracting environments, it is generally possible to allocate ordinary income or loss, and capital gain or loss differently across investors to best match their tax clienteles. In the years preceding the Tax Reform Act of 1986, firms with low marginal tax rates had incentives to form partnerships to share their losses with high marginal tax rate investors; Shevlin [1987] found evidence of this activity. However, the 1986 Act limited deductibility of losses from passive investments to only against passive income and Beatty et al. [1995] find mixed support for the tax argument as they find a negative tax rate effect even after the 1986 Act. Currently, synthetic leases that are used widely provide an excellent example of the “best of both worlds” treatment that allows a sponsor to treat the lease of an asset from an SPE as a capital lease for tax purposes and as an operating lease for financial reporting purposes. Because capital lease treatment results in greater tax benefits, firms with high tax rates have stronger incentives to set up SPEs for synthetic leases. Consistent with this argument, Mills and Newberry [2005] find evidence that firms with larger book-tax difference in interest expense have fewer NOLs or have higher marginal tax rates. We use MTR based on Graham’s [1996] firm-specific simulation of the likelihood of incurring and utilizing net operating loss carryforwards, and expect a positive coefficient on MTR. Optimal risk clienteles. Sponsors also use SPEs to meet demands from lenders who seek to invest in homogenized risks with collateralized assets that are legally isolated from the firm’s

16

general creditors. Therefore, when a sponsor transfers an asset or class of assets with homogenized risks to a legally isolated SPE, the SPE may become attractive to an investment clientele seeking certain parameters of risk and return. As Beatty et al. [1995] discuss, it is difficult to predict how holding risky assets and liabilities will impact SPE creation. Firms may transfer risky assets to SPEs to reduce investors’ perception of risk. Conversely, some firms may transfer less-risky assets to SPEs to raise capital from risk-averse investors who otherwise might avoid investing in the firm. We proxy management’s desire to shift risky assets and liabilities to SPEs with RISK estimated by a two-step process (see Beatty et al.). First, standard deviations of all CRSP firms’ daily stock returns are ranked and sorted into deciles (0 for the lowest risk, 9 for the highest risk). Second, each firm in the sample receives a value for RISK between 0 and 9 depending on the portfolio that best fits the firm’s stock return standard deviation. 4.2.

FINANCIAL REPORTING DETERMINANTS Debt covenants specify in accounting terms the minimum financial requirements

necessary for a firm to avoid technical default in lending arrangements. Dichev and Skinner [2002] document large-sample evidence based on detailed debt terms reported in DealScan that firms advance future income into the current period when they are close to violating accountingbased covenants. Dichev and Skinner also determine that (1) leverage is a noisy but fairly effective proxy for the tightness of debt covenant restrictions, consistent with evidence in Press and Weintrop [1990] and Duke and Hunt [1990], and (2) the interest coverage ratio is a commonly employed covenant. We proxy the closeness of firms to debt covenant restrictions with LEV, the ratio of total debt to total assets, and INTCOV, pre-interest operating income divided by interest expense. We expect SPE use to be increasing in leverage and decreasing in interest coverage.

17

The bonus plan hypothesis argues that managers adjust reported earnings when their compensation depends on performance contracts that are explicitly or implicitly based on reported earnings. Healy [1985] offers early evidence supporting the hypothesis. Bushman and Smith [2001] review this literature, summarizing multiple studies that buttress the general tenets of Healy’s original research. Ittner, Larcker and Rajan [1997] report that variants of accounting earnings are the most common financial metrics used in determining manager bonuses. We proxy earnings-based performance incentives with BONUS, defined as the ratio of the CEO’s bonus to total current compensation (salary plus bonus) and expect that firms’ SPE sponsorship is increasing in the percentage of the CEO’s compensation from bonuses. Dechow, Sloan and Sweeney [1996] find that companies committing accounting fraud to manipulate earnings upward are more likely to issue capital in the future and so conclude that the desire to raise external financing at low cost is an important motivation for earnings manipulation. Consistent with this argument, Teoh, Welch and Wong [1998] and Teoh, Wong and Rao [1998] find that firms issuing capital report unusually large positive accruals in the year preceding capital issuances, but that they experience lower stock returns in subsequent periods. We include STOCKISS, the value of net stock issuances deflated by average total assets, and DEBTISS, the net increase in debt deflated by average total assets, to capture firms’ incentives to manage their financial statements prior to raising capital externally. We expect SPE use to be increasing in both variables. 4.3.

CORPORATE GOVERNANCE VARIABLES The Business Roundtable [2002] and others (e.g., Lipton and Lorsch [1992]) advise that

sound governance necessitates that the majority of board members be independent of management. The logic is that outside or independent directors monitor managers to increase

18

agency alignment between shareholders (i.e., principals) and managers (i.e., agents). Brickley, Coles and Terry [1994] and Byrd and Hickman [1992] determine that greater independence among directors increases board monitoring effectiveness. Larcker, Richardson and Tuna [2004] uncover evidence that analyst stock recommendations and bond ratings are lower among firms with strong insiders, a construct that depends partially on a measure of the proportion of independent directors. We proxy board member independence directly with the percentage of independent directors on the board, DIRIND, and expect that SPE use is decreasing in DIRIND. The Business Roundtable [2002] also suggests that outside directors ought to be “incentivized to focus on long-term stockholder value” by including firm equity as part of board members’ compensation. Klein [1998] finds evidence of positive relation between shareholder value and stock ownership by outside directors who serve on the investment and finance committees. We include INDSH, the sum of the percent of voting common shares held by independent directors, and expect that SPE use is decreasing in the independent directors’ proportion of share ownership. Directors serving on “too many” boards risk the possibility that their multiple-firm service can compromise the ability to perform well for any one of the boards (Business Roundtable [2002], Lipton and Lorsch [1992]). Core, Holthausen and Larcker [1999] find that firms with higher proportions of busy directors provide higher compensation to chief executive officers. Larcker, Richardson and Tuna [2004] find that stock analyst recommendations are lower among firms with greater proportions of busy directors. Conversely, however, Ferris, Jagannathan, and Pritchard [2003] find no evidence supporting the hypothesis that directors serving on multiple boards shirk their responsibilities or that their firms are more likely to be

19

named in securities fraud lawsuits. Similar to Core, Holthausen and Larcker, we include, BUSY, defined as the percentage of directors serving on more than two other boards. 4.4.

OTHER CONTROLS SIZE is the sum of the market value of common equity and the book value of preferred

equity and debt. As the SEC [2005, p.2] notes, “SPEs are likely to be disproportionately concentrated in the very largest issuers” mainly because larger firms likely possess greater technical expertise to handle the complexity of structured finance arrangements and, hence, more able to utilize SPEs as a means of raising capital (i.e., we predict a positive coefficient for SIZE). INDU_PERC is a variable that controls for industry membership using the Fama and French [1997] industry categories. The variable is defined as the percentage of firms with SPEs within each industry for each year; positive coefficients are expected because setting up SPEs is likely influenced by industry practices. Finally, year dummies (YEAR) are designed to capture other systematic economy-wide effects varying over time.

5.

Results for the determinants of SPE use

5.1.

DESCRIPTIVE STATISTICS AND UNIVARIATE TESTS Table 5 provides a variety of descriptive information about the explanatory variables used

in the regressions. Panel A presents descriptive statistics. Panel B reports results of univariate tests of differences in means and medians between firms with zero SPEs and firms with at least one SPE. Panel B also presents mean and median values for each variable partitioned by categories of the number of SPEs. Panel C reports pairwise correlations. Panel A shows that the median (mean) supply of funds (FUNDS) is 0.002 (-0.032), indicating that most firms produce positive internal cash flow but that a minority of firms with

20

substantial negative flows skew the distribution to the left. Similarly, while median debt due within a year (CLTD) is 2.3 percent of total assets, the mean is 5.4 percent, suggesting that a few observations have substantial debt due within a year. The mean (median) marginal tax rate, MTR, is 30 (35) percent across the sample period. For the financial reporting variables, the sample firms’ mean and median ratio of debt to total assets, LEV, is about 59 percent and mean (median) interest coverage ratio,INTCOV, is about 34 (5) times. Bonus as a percentage of total current compensation (salary plus bonus) is an average (median) of 37 (40) percent. The median sample firm has zero net issuances of debt, DEBTISS, and stock, STOCKISS. Among governance variables, on average two-thirds of sample firms’ board of directors are independent (DIRIND) and about 10 percent of their board members are considered BUSY. Panel B of Table 5 compares explanatory variables between firms with and without SPEs and shows that many of the differences are consistent with our predictions. Relative to firms with no SPEs, firms reporting at least one SPE have higher marginal tax rate, leverage, lower interest coverage, and their CEO bonuses represent a greater portion of total current compensation. Firms with SPEs are also larger and they are in industries with a higher proportion of firms using SPEs relative to their counterpart firms with no SPEs. However, firms with SPEs indicate that a smaller proportion of their debt is due in the following year, reflecting lower renegotiation needs, which is inconsistent with predictions. Panel C of Table 5 shows that the pairwise correlations between the explanatory variables (Spearman [Pearson] correlation coefficients are above [below] the diagonal) are generally low (less than 0.30), except for some of the correlations with firm size. Some of the larger (Spearman) correlations with firm size are with MTR (0.35), RISK (-0.54), LEV (0.37) and BONUS (0.32). RISK and MTR are also correlated at -0.38, and LEV and INTCOV are correlated

21

at -0.50. Pearson correlations are of similar, but generally smaller, magnitudes. These statistics suggest that multicollinearity is not a concern in interpreting the regression results. 5.2.

REGRESSION RESULTS OF SPE USE Table 6 presents the results of the four Tobit regression models of the number of SPEs

reported in Exhibit 21 of Form 10K, with the two-tailed chi-square statistic shown in parentheses. All four models are significant overall (the likelihood ratio chi square is significant at p