Do operating leases expand credit capacity? Evidence from borrowing costs and credit ratings

    Do operating leases expand credit capacity? Evidence from borrowing costs and credit ratings Steve C. Lim, Steven C. Mann, Vassil T. Mihov PII: DOI: Reference:

S0929-1199(16)30217-6 doi: 10.1016/j.jcorpfin.2016.10.015 CORFIN 1112

To appear in:

Journal of Corporate Finance

Received date: Revised date: Accepted date:

31 July 2015 24 October 2016 25 October 2016

Please cite this article as: Lim, Steve C., Mann, Steven C., Mihov, Vassil T., Do operating leases expand credit capacity? Evidence from borrowing costs and credit ratings, Journal of Corporate Finance (2016), doi: 10.1016/j.jcorpfin.2016.10.015

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Do operating leases expand credit capacity? Evidence from borrowing costs and credit ratings

AC CE P

TE

D

MA

NU

SC

RI

PT

Steve C. Lima, Steven C. Mannb, Vassil T. Mihov c,*

a

Department of Accounting, Neeley School of Business, Campus Box 298530, Texas Christian University, Fort Worth, TX 76129 b Department of Finance, Neeley School of Business, Campus Box 298530, Texas Christian University, Fort Worth, TX 76129 c Department of Finance, Neeley School of Business, Campus Box 298530, Texas Christian University, Fort Worth, TX 76129 * Corresponding author, email: [email protected]

ACCEPTED MANUSCRIPT Abstract

RI

PT

We document that borrowing costs and credit ratings are less sensitive to off-balance sheet lease financing than to on-balance sheet debt financing, particularly for firms that are financially constrained and firms that have limited ability to use tax shields. This evidence is consistent with theoretical predictions based on tax benefits as well as bankruptcy costs. Our evidence on borrowing costs and credit ratings suggest that credit markets treat operating leases differently from balance sheet debt. Consistent with this interpretation, we document that firms closer to ratings borderlines lease more, particularly around the investment grade borderline.

SC

Keywords: Operating leases, Off-balance sheet financing, Cost of capital, Credit capacity

AC CE P

TE

D

MA

NU

JEL classification: G32, M41

ACCEPTED MANUSCRIPT 1.

Introduction Operating leases are the most common and important source of off-balance sheet

PT

financing. The Morning Ledger from CFO Journal estimates on August 11, 2014 that operating

RI

leases represent about $2 trillion in off-balance sheet financing. On February 25, 2016, the FASB issued a new standard, Leases (ASC 842) requiring companies to add long-term operating

SC

leases to the balance sheet. The new accounting standard will dramatically boost reported

NU

leverage for many firms.

Prior studies document that lessees incur significant transaction costs to obtain off-

MA

balance sheet treatment of lease contracts (Imhoff and Thomas 1988; Zechman 2010; Schallheim, Wells, and Whitby 2013). We investigate potential benefits of leases in expanding or

TE

D

preserving credit capacity. If leases displace less than an equivalent amount of debt, then firms may be able to use leases to expand credit capacity.

AC CE P

Specifically, we document that borrowing costs and credit ratings are less sensitive to offbalance sheet lease obligations than to on-balance sheet debt indicating that leasing is advantageous in the sense of lowering borrowing costs. This effect is more pronounced for financially constrained firms and for firms with low marginal tax rates. Our findings support predictions of leasing models regarding the minimization of expected bankruptcy costs (Eisfeldt and Rampini, 2009) and the sharing of tax shields (Lewis and Schallheim, 1992). Our evidence on the differential effect of leases compared to debt on cost of borrowing and credit ratings extends and complements studies documenting that lease obligations are incorporated in bond yields, bank loan rates and credit ratings (Lim et al., 2003, Bratten et al., 2013, Alatamuro et al., 2014). Our evidence is also consistent with Schallheim et al. (2013) who

1

ACCEPTED MANUSCRIPT provide evidence suggesting that operating leases expand credit capacity in sale-and-leaseback transactions.

PT

Consistent with the idea that leases allow some firms to expand credit capacity, we provide evidence that firms use operating leases to manage credit ratings. Graham and Harvey

RI

(2001) provide survey evidence that “the two most important factors affecting debt policy are

SC

financial flexibility and a good credit rating” (p. 189). Alissa et al. (2013) and Jung et al. (2013)

NU

document that firms use accounting discretion to affect credit ratings. We document that even though credit agencies take leases explicitly into consideration, credit ratings are less sensitive to

MA

leasing than to debt.

If off-balance sheet lease financing affects credit risk less than on-balance sheet debt

TE

D

financing, we expect firms on ratings borderlines to lease more to preserve or enhance their credit ratings. We confirm this expectation, as we find that firms close to ratings borderlines are

AC CE P

more likely to lease, particularly those firms near the investment grade borderline. The evidence on borrowing costs, credit ratings, and lease usage is robust to different methods of estimating the lease obligations, as well as an alternative methodology based on the “abnormal” component of leasing used by Cornaggia, Franzen, and Simin (2013). The remainder of the paper is structured as follows. Section 2 explains the linkage between borrowing costs, credit ratings, and debt capacity. Section 3 explains our methodology and data sources. Section 4 examines the impact of leasing on bank borrowing costs, and yields on bonds. Section 5 examines the relationship between leasing and credit ratings. Section 6 concludes the paper.

2

ACCEPTED MANUSCRIPT 2. Borrowing costs, credit ratings, and debt capacity

PT

Traditionally, finance theory assumes that lease obligations substitute for debt in the capital structure by using limited debt capacity. If leasing simply replaces debt in the capital

RI

structure, then why are firms willing to incur significant transactions costs required for leases to

SC

meet accounting requirements for operating lease treatment? In other words, what specific benefit does off-balance sheet lease financing create for the lessee?

NU

We define credit capacity as the optimal amount of combined balance sheet debt and off-

MA

balance sheet lease obligations. If leases displace less than an equivalent amount of debt, then firms use leases to expand credit capacity. The literature has three alternative explanations for

D

increased credit capacity associated with lease financing. They are the minimization of agency

TE

costs in Smith and Wakeman (1985), the minimization of bankruptcy costs in Eisfeldt and

AC CE P

Rampini (2009), and the sharing of tax deduction benefits in Lewis and Schallheim (1992). Empirical evidence indicates a positive correlation between debt and lease usage, e.g., Ang and Peterson (1984), Eisfeldt and Rampini (2009), Rauh and Sufi (2010), Cornaggia et al. (2013), and Schallheim et al. (2013). But as Lewis and Schallheim (1992) point out, the positive correlation may simply indicate that firms with greater credit capacity and requirements for debt financing use balance sheet debt and leases interchangeably. Other papers examine changes in reported capital over time and find that on average, leasing substitutes for debt, but not dollar for dollar, e.g., Marston and Harris (1988) and Yan (2006). Borrowing costs and credit ratings reflect the size and utilization of a firm’s debt capacity. Myers et al. (1976) model lease valuation and identify the fraction of a dollar of debt displaced by a dollar of present value of lease obligations (  ). In their framework, if leases and debt are perfect substitutes,  equals1, otherwise they are imperfect substitutes and 0    1 3

ACCEPTED MANUSCRIPT implies leases expand debt capacity. In what they describe as “the leasing puzzle”, Ang and Peterson (1984) find that debt and leases appear to be complements instead of substitutes, as they

PT

observe   0  Myers et al. (1976) note that, from the lessee perspective,  should be less than

RI

1 due to sharing of tax benefits. When tax shields are transferred from a borrower/lessee with limited ability to use tax shields to the firm with greater ability to use the deductions (the

SC

lender/lessor),  should be less than 1. But Myers et al. (1976) leave the question of the degree

NU

of substitutability as an open issue.

Subsequent work has proposed that leases do not fully substitute for debt, but displace

MA

less than a dollar of debt for a dollar of leasing. Lewis and Schallheim (1992) focus on sharing tax benefits and show that leasing does not necessarily displace debt dollar for dollar. They

TE

D

demonstrate a theoretical possibility that leases do not displace any debt at all such that credit capacity expands by more than the amount of leasing.

AC CE P

Eisfeldt and Rempini (2009) suggest that leasing minimizes bankruptcy costs because of the fully collateralized nature of leasing.

Operating leases have relatively straightforward

clauses for assignment of collateral and leased assets are more easily repossessed. Eisfeldt and Rampini (2009) formally model this concept, and show that, by having fewer creditors (and fewer assets) involved in bankruptcy claims, expected bankruptcy costs to creditors/lessors can be reduced. By reducing expected bankruptcy costs, leasing can increase expected recovery in the case of default. They show that the increased expected recovery upon default created by financing a portion of assets with fully collateralized operating lease should decrease the cost of debt financing compared to financing assets exclusively with balance sheet debt. The theoretical explanations for increased credit capacity all predict that replacing debt with an equivalent amount of leasing should increase the total expected cash flow available to

4

ACCEPTED MANUSCRIPT service and repay debt. Holding leverage constant, the increased available cash flows should lower the cost of borrowing. Alternatively, the increased expected cash flow associated with

PT

leasing could be used to support a higher level of leverage without increasing the cost of debt, so

RI

that increased credit capacity would be reflected in higher leverage. If operating lease financing increases credit capacity by lowering borrowing costs, then leases should have a lower impact on

SC

borrowing costs than an equivalent amount of on-balance sheet debt.

NU

In the case of borrowing costs, Modigliani and Miller (1958) propose that lenders require higher yields as borrowers increase the proportion of fixed claims (debt plus leases) in the capital

MA

structure, since “the further a claimant stands from the head of the line at payoff time, the riskier the claim.” (Miller, 1991, p. 482). In a perfect world, the overall cost of debt will be constant, as

TE

D

credit is safe. In a world with costly financial distress, the overall cost of borrowing will increase as a function of the amount of credit (relative to equity). If the firm is financed with unsecured

AC CE P

debt only, its cost of debt will increase faster than if it is financed with a mix of unsecured debt and safer, secured obligations such as leasing. Introducing a safer, secured obligation to the credit mix will reduce the overall borrowing cost, holding constant the level of credit. This result has two equivalent implications. First, transforming the credit structure from unsecured debt to a mix of secured and unsecured debt lowers the cost of borrowing. Second, given a certain level of borrowing cost, RD* (dictated by a desired credit rating, for example) the firm can borrow more using a mix of secured and unsecured debt than using only unsecured debt. We illustrate these two equivalent implications in Figure 1.

5

ACCEPTED MANUSCRIPT In this paper we do not explicitly model debt capacity. Instead, we assume that debt capacity is associated with borrowing costs and credit ratings in that, ceteris paribus, borrowing

PT

costs rise and credit ratings drop as leverage increases. We assume the cost of borrowing for any

balance sheet debt; present value of off-balance sheet lease obligations; credit factors other than debt and leases; total assets, including the present value of the lease obligations; monotonic function with positive first derivative f’; monotonic function with positive first derivative g’; impact of other credit factors on borrowing costs.

NU

= = = = = = =

MA

where D L Z TA f(D/TA) g(L/TA) h(Z)

SC

Cost of borrowing = f(D/TA) + g(L/TA) + h(Z),

RI

particular firm can be represented as:

D

In this parsimonious model of borrowing costs, we assume that borrowing costs increase

TE

with leverage, whether on-balance sheet or off-balance sheet, ceteris paribus. We extend similar logic to analysis of credit ratings, positing that credit rating drops as leverage increases.

AC CE P

If leasing fully substitutes for balance sheet debt (  1), so that leasing fully displaces debt on a dollar-per-dollar basis, then f '  D / TA  g '  L / TA , so that lenders consider leasing and balance sheet debt equivalently in terms of using debt capacity. On the other hand, if leases substitute less than debts (   1), then we would expect f '  D / TA  g '  L / TA such that balance sheet debt uses more debt capacity than leasing.

6

ACCEPTED MANUSCRIPT 3. Data and methodology 3.1. Estimating the value of the leasing obligation

PT

We estimate the value of operating lease obligations with three different methods. The first is S&P’s present value method, which capitalizes the present value of minimum lease

RI

payments for five years plus the estimated remaining years in the “thereafter” value. We

The second is Moody’s multiple method and it

NU

fifth year minimum lease commitment.

SC

estimate the minimum payments for the remaining years by dividing the thereafter value by the

capitalizes operating leases by selecting the higher value between S&P’s and 8 times current year

MA

rent expense. The Moody’s debt equivalent amount is always greater than or equal to S&P because S&P’s goal is to capture only the contractual obligation while Moody’s attempts to

TE

payments are perpetual.

D

capture the value of the entire obligation. The Moody’s method essentially assumes lease

AC CE P

We also use a third method (“LMM”) of operating lease capitalization developed by Lim, Mann and Mihov (2003), based on Miller and Upton (1976). 1 The LMM approach capitalizes lease payments using a multiple driven by estimated interest and depreciation.

The lease

payment ( P ) is comprised of interest charges plus depreciation charges. The LMM method computes the depreciation adjusted perpetuity estimate ( V ) using the average of current year rental expense and the next year minimum lease payment as a proxy for lease payment ( P ). The depreciation on the leased assets is V / N , where N is the estimated average useful life of leased assets (measured by PPE divided by annual depreciation expense). Note that the cost of the depreciation charge to the lessee is reduced to V / N  * 1    due to the tax benefits to the lessor, where  is statutory maximum tax rate (35%). The interest is k *V , where k is the cost

1

The LMM method or its close variation is used by Yan (2006), and Eisfeldt and Rampini (2009).

7

ACCEPTED MANUSCRIPT of debt capital. Combining the two components of the lease payments (interest and depreciation) produces the following expression:

 N  1  t   V k   1 N  1  t 

PT

P  k *V  V

1 t N 

.

SC



k 1

P

NU

V

RI

Rearranging the expression generates the LMM estimated value of the lease obligation:

3.2. Data

MA

Our sample consists of annual North America Xpressfeed COMPUSTAT firms with SIC codes between 2000 and 5999 during the period 1995 to 2011 with minimum lease payments

D

data available. The final sample has 31,339 firm/year observations on 5,378 firms.

TE

We use time varying risk and maturity adjusted discount rates (k) for estimation of lease

AC CE P

obligations. We obtain annual yields on five-year maturity corporate industrials across the rating spectrum from Bloomberg.2 The yields are matched with each firm/year bond ratings, which are obtained from historical S&P long-term issuer credit ratings from COMPUSTAT. For unrated firms, we first estimate a synthetic rating based on their size and coverage ratio and then apply the matched yield.

We obtain data on new debt issues from Thompson Financial’s New Issues database (SDC). We obtain data on the issue date, the identity and characteristics of the borrower, and various characteristics of the bond issue, such as proceeds in nominal dollars, maturity, yield to maturity (YTM) at issuance, credit rating, and call or put features. From the Federal Reserve Bank of Saint Louis’ FRED database we obtain time series of the monthly yields on Treasury securities with maturities of six months, one year, two years, three years, five years, seven years, 2

We also replicated the analysis using yields for ten-year maturity bonds, with similar (unreported) results.

8

ACCEPTED MANUSCRIPT ten years, 20 years, and 30 years. In our analysis of the impact of leases on bond borrowing costs, we measure the yield spread as the difference between the issue YTM and the yield on a

PT

maturity-matched Treasury.

database.

RI

We also obtain data on bank loans reported by Loan Pricing Corporation Dealscan We obtain data on the loan date, the identity of the borrower, and various

SC

characteristics of the loan facility, such as facility amount, maturity, covenants, and the all-in-

NU

drawn spread. 3.3. Descriptive Statistics

MA

We compare the relative magnitude of off-balance sheet operating lease to on-balance sheet total debt represented by the sum of book value debt and capitalized leases. In Table 1, we

TE

D

present mean and median values for debt and operating lease estimates relative to total assets. Balance sheet debt comprises, on average, about 32% of total assets (median value 26%).

AC CE P

Operating leases represent about 10% of total assets (median 4%) using the S&P method, and 25% (median 12%) using the Moody’s multiple method, and 21% (median 10%) using the LMM method.

Firms in different ratings use different levels of operating lease financing. Investment grade firms use less debt and less leasing compared to sub-investment grade firms. Rated firms, however, use more debt compared to firms that are not rated (on average, 38% vs. 27%) but less leasing (8% vs. 11.2% using the S&P method, 19% vs. 29% using the Moody’s method, and 18% vs. 24% using the LMM method). We also observe differences in leasing and debt use across the firms that issue bonds during our sample period, and those that did not (as reported by the SDC database). Bond issuers use more debt (35% vs. 30.5%) but less leasing (7.9% vs. 10.5% using the S&P method) compared to those firms that did not issue bonds. Firms that

9

ACCEPTED MANUSCRIPT obtain bank loans recorded by Dealscan also use more debt (32.1% vs. 30.5%) but less leasing

PT

(24% vs. 27% using the Moody’s method) compared to firms that did not obtain bank loans.

4. The impact of operating leases on borrowing costs

RI

In this section we examine the effect of operating leases and debt on the cost of

SC

borrowing with new bonds first and bank loans next. 4.1. New Bank Loans

NU

In Table 2 we present the results on the cost of new bank loans. We examine

MA

determinants of borrowing costs measured as the “All-in-Drawn Spread” on new bank loan facilities reported by Dealscan. The “All-in-Drawn Spread” is the spread over LIBOR, including

D

interest and fees, in basis points, for a dollar drawn from the loan facility. Dealscan re-calculates

TE

the spread for loans that are not based on LIBOR. We control for firm and/or loan characteristics

AC CE P

that affect borrowing costs such as firm total assets, profitability (measured by EBITDA/TA), asset tangibility (measured by Net PPE/TA), issue size, and loan maturity. In order to account for the significant time variation in the cost of bank borrowing due to macroeconomic factors, we use the TED spread, defined as the spread between 3-month LIBOR based on US dollars and 3month Treasury bill, as reported by St. Louis FRED database.3 We recognize that there is a considerable heterogeneity in loan contract covenants, and examine the Dealscan database for restrictive covenants associated with each loan. Following Demerjian (2011), we classify covenants into three categories: 1) earnings-based covenants, including requirements for interest coverage, fixed charge coverage, debt-to-EBITDA, and minimum level of EBITDA; 2) balance-sheet based covenants, including net worth, leverage

3

We winsorize the following variables at the 1% and 99% to limit the influence of extreme observations: debt to total assets, lease obligations to total assets, EBITDA to total assets, net PPE to total assets, and amount issued to total assets.

10

ACCEPTED MANUSCRIPT (including debt-to-equity, debt-to-assets, equity-to-assets, and debt-to-net worth), current ratio, and quick ratio; and 3) investment-based covenants including restrictions on capital expenditures

PT

or long-term investment to net worth. We create indicator variables for the presence of each of

RI

these categories of covenants.

Leases might mitigate the effect of some covenant restrictions. For example, a number of

SC

loans have book leverage covenants, such as debt-to-book equity, debt-to-book value of assets

NU

(or net worth), restricting future borrowing to protect the existing lender. In this case a firm might turn to operating leases to regain or preserve debt capacity. We also create a dummy

MA

variable equal to 1 if the loan is senior, 0 otherwise. (Senior loans have priority to junior/subordinated or mezzanine debt, and are often secured, while subordinated loans are

TE

D

typically unsecured).

We find that larger size firms and more profitable firms have lower borrowing spreads,

AC CE P

larger size loans have lower spreads, and loans with longer maturity have higher spreads. Surprisingly, we find that firms with higher tangible assets have higher spreads.4 We find that balance sheet based covenants have a significant negative effect on loan spreads, while earnings based covenants have a smaller negative effect (that is sometimes not statistically significant at the 10% level).

Investment based restrictions are associated with higher yields, possibly

indicating greater degree of risk shifting/moral hazard in such loans. The seniority dummy is strongly negatively associated with loan spreads. We document that the cost of new bank loans increases in both existing debt and operating leases. However, balance sheet debt has much higher marginal effect on the cost of new borrowing than existing leases. For each additional percent of existing debt relative to total

4

Plumlee et al. (2015) document that banks incorporate private information about borrowers’ forthcoming patents into bank loan spread pricing.

11

ACCEPTED MANUSCRIPT assets, borrowing costs increase by 1.47 basis points, while for each additional percent of leases relative to total assets, the cost of borrowing increases by 0.38 basis points using the S&P

PT

method (1.47 vs. 0.25 using the Moody’s method, and 1.48 vs. 0.12 using the LMM method).

RI

We reject the null that the coefficients on debt and leases are equal to each other at statistical levels less than 0.001 in all specifications.

SC

Even though both debt and leases are measured in the same units (i.e., in constant dollars Regression slope

NU

and scaled by total assets), we also report the standardized coefficients.

estimates depend on the level of regressors in that larger variables will result in smaller

MA

coefficients. The slope coefficient of each standardized estimate is interpreted as change in the yield spread, in standard deviations, resulting from one standard deviation change in the

TE

D

independent variables. Thus, one standard deviation change in debt-to-assets will result in .24 standard deviation changes in the borrowing spread compared to .04 standard deviation change

AC CE P

as a result of leases measured by the S&P method, .24 vs. .05 when comparing debt-to-leases measured by Moody’s method, and .24 vs. .02 using the LMM method. The results are consistent across specifications and different methods for estimating the value of operating leases. For robustness, we also employ a technique developed by Cornaggia, Franzen, and Simin (2013). They decompose the level of leasing into a predicted and “abnormal” component by regressing leases against independent variables suggested by Graham, Lemmon, and Schallheim (1998).

The variables include before-interest marginal tax rate, expected cost of distress,

modified Altman’s Z-score, a dummy variable for whether owners' equity is negative, market-tobook ratio, collateral, firm size, as well as industry dummy variables and time period indicators. 5 We report these results in Table 2, Panel B.

5

The decomposition causes us to lose about one half of the observations for the bank loans subsample due to the data requirements on R&D and advertising expenses used to calculate expected cost of distress. We note, however,

12

ACCEPTED MANUSCRIPT In all specifications, we observe that the “abnormal” component of leases has a lower coefficient than that of debt, i.e., a dollar of discretionary use of leases affects the cost of new

PT

debt by less than a dollar of debt. In unreported regressions, we included specifications with

RI

both the predicted and abnormal components of leases. Both components had lower coefficients than debt, and the abnormal component had the lowest effect on cost of debt. We note that in the

SC

last regression of Table 2, Panel B, the coefficient of the “abnormal” leasing using LMM method

NU

is not statistically significantly different from zero. However, the coefficient of the predicted

MA

amount is positive, but smaller than that of debt, in the unreported regression. 4.2. New Bond Issues

In Table 3 we extend the analysis to examine the yield-to-maturity (YTM) on new bond

TE

D

issues as reported by SDC. We subtract from the bond YTM the yield of a Treasury with similar maturity, measured in percentage points. In addition to debt and leases, we control for other

AC CE P

factors affecting bond pricing such as firm total assets, profitability (measured by EBITDA/TA), asset tangibility (measured by Net PPE/TA), and issue size. We control for the significant timevariation of the spread itself, resulting from changing by macroeconomic conditions by including the difference between the yields on the Moody’s seasoned BAA corporate bonds and the 10year constant maturity Treasury, as reported by the St. Louis FRED database. As with bank loans, there is cross-sectional variation in the characteristics of the bond contracts. We account for whether the bond is callable or puttable, as well as for the bond maturity (in the construction of the yield variable, which is matched to a Treasury security with similar maturity). However, we do not account for bond covenants due to lack of access to such

that the regression specifications are remarkably similar in terms of adjusted R-square and overall coefficient estimates. The loss of observations is substantially smaller for the bond subsample.

13

ACCEPTED MANUSCRIPT data. We consider the lack of covenant data for the bond subsample to be a limitation of our study, and ask the reader consider it in interpreting our results for the bond subsample. 6

PT

We find that larger firms and more profitable firms have lower yield spreads, firms that

RI

borrow more have higher yield spreads, callable issues are costlier and puttable issues are less costly. Both existing debt and leases affect positively the cost of borrowing. However, the

SC

effect of balance sheet debt and leases on the cost of new borrowing varies across lease

NU

capitalization methods. In the case of the S&P method, even though the debt estimate, 2.05, appears higher than that of leases, 1.80, we cannot reject the null of their equality with an F-

MA

value of 0.95. This is consistent with findings in Bratten et al. (2013). If we use the Moody’s method, the respective coefficients are 2.05 vs. 1.23, and they are significantly different from

TE

D

each other at less than 0.001 level, with an F-value 19.11. Similarly, with the LMM method, the respective coefficients are 2.11 for debt, and 1.03 for leases, also significantly different at the

AC CE P

0.001 level.

In terms of standardized coefficients, one standard deviation change in debt-to-assets ratio will result in .19 standard deviation changes in the yield spread compared to .11 standard deviation change in the leases-to-assets ratio measured by the S&P method (0.19 vs. 0.13 when comparing debt-to-leases measured by Moody’s method, and 0.20 vs. 0.11 using the LMM method).

6

While our paper would be stronger if we had data on bond covenants, we believe that our current results are robust to the omission of covenants. Bond covenants exhibit much less cross-sectional variation compared to bank loan covenants. For example, Reisel (2014) documents that more than 90% of the bonds have a restriction on financial activities. In comparison, only 62% of the bank loans in our sample have financial covenants (of any kind). As prior theoretical and empirical research (for example, Rajan and Winton, 1995, Denis and Wang, 2014, among others) has argued, banks are provided with stronger incentives to monitor by inclusion of covenants, and are better suited to enforce or renegotiate covenant violations than public bondholders. In other words, covenants should have a stronger effect on the bank loan subsample than the bond subsample, and we are comforted that our main results remain unchanged for the bank loan subsample when we include covenants.

14

ACCEPTED MANUSCRIPT Overall, we document that, as with bank loans, borrowing costs for newly issued bonds increase as firms add debt, whether on- or off-balance sheet, but are more sensitive to balance

PT

sheet debt than to off-balance sheet lease financing for those methods that account for the

RI

potential of lease renewal (i.e., Moody’s method and LMM method) as opposed to measuring the minimum contractual obligation only (i.e., S&P method). Once again, for robustness, in Table 3,

SC

Panel B, we used the lease decomposition suggested by Cornaggia, Franzen, and Simin (2013).

NU

As in Table 2, Panel B, in all specifications, the “abnormal” component of leases has lower coefficient than that of debt, including that for the S&P method; in unreported regressions, both

MA

the predicted and abnormal components of leases had lower coefficients than debt, but the

D

abnormal component had the lowest effect on cost of debt.

AC CE P

5.1. Financial Constraints

TE

5. Cross-sectional variation in the impacts of operating lease on borrowing costs

We extend the results reported in Table 3 (the impact of on-balance sheet debt and offbalance sheet lease financing on borrowing costs) by introducing variables that proxy for credit constraints. First, we use a proxy for financial constraints based on a measure for free cash flow (FCF) following Barry et al. (2008) in Panel A of Table 4. We calculate the income before extraordinary items but after tax plus depreciation minus capital expenditures scaled by total assets, and divide the sample into high FCF and low FCF firms around the median. Then, we create orthogonal variables for debt and leases based on whether the firm had a ratio of FCF to total assets above or below the sample median. Specifically, TD/TA for high FCF firms is equal to TD/TA if the firm had a ratio of FCF to total assets above or equal to the median, 0 otherwise; TD/TA for low FCF firms is equal to TD/TA if the firm had a ratio of FCF to total assets below the median, 0 otherwise. Similarly, Leases/TA for high FCF firms is equal to Leases/TA if the 15

ACCEPTED MANUSCRIPT firm had a ratio of FCF to total assets above or equal to the median, 0 otherwise; Leases/TA for low FCF firms is equal to Leases/TA if the firm had a ratio of FCF to total assets below the

PT

median, 0 otherwise.

RI

For firms that are in the low FCF category, the coefficient for debt is 2.54 and for leases 1.42, using the S&P method (the two coefficient are statistically different at 0.001 level with F-

SC

value 10.87). For firms that are in the high FCF category, the effect of debt is 1.53, while that

NU

for leases is 2.09 (with a p-value for difference of .115). In other words, for firms that are more likely to be constrained, leasing affects the cost of new borrowing less than on-balance sheet debt The same inferences are obtained using the

MA

compared to firms that are less constrained. Moody’s and LMM methods.

TE

D

In Panel B, we create similar orthogonal variables (rated firms and unrated firms) based on whether the firm had prior rated debt as indicated in COMPUSTAT. Faulkender and Petersen

AC CE P

(2006) indicate that non-rated firms have less access to debt capital that results in overall leverage that is lower by as much as 35% compared to rated firms. For firms that have not issued rated debt, the coefficient for debt is 3.16, compared to 1.68 for leases (the two coefficients are statistically different at 0.068 level, F-value 3.34), using the S&P method. For firms that have previously issued rated debt, the coefficient of debt is 1.95 vs. 1.83 for leases (pvalue 0.638), statistically and economically similar. The results are even more pronounced using the Moody’s or the LMM method. Thus, for firms that have not issued rated debt before, the difference between the effect of debt and leases on the pricing of debt is greater, consistent with the idea of leasing extending borrowing capacity for constrained firms. We do observe in Table 1 that non-rated firms use less debt and more leases compared to rated firms.

16

ACCEPTED MANUSCRIPT Finally, we examine the ratio on interest expense to total assets following van Binsbergen et al. (2010) with two orthogonal variables (low interest obligation firms and high interest

PT

obligation firms). We split the sample firms around the median value into those that have low

RI

interest obligations (less likely to be constrained) and high interest obligations (more likely to be constrained). In Panel C of Table 4, for firms in the high interest obligation category, the

SC

coefficient for debt (1.94) is significantly higher than that for leases (1.06) using the Moody’s

NU

method. For firms in the low interest obligation category the difference is insignificant. The differential effect of leases and debt for constrained vs. unconstrained firms holds true if we

MA

consider the LMM specification, but not for the S&P specification. In unreported regressions, we substituted the use of leases with its abnormal components.

TE

D

Qualitatively, the comparative results for the financially constrained and unconstrained are robust to this alternative specification. Overall, we observe that financially constrained firms

AC CE P

benefit more from lower marginal effect of leasing on the cost of new debt vis-à-vis on-balance sheet debt financing.

5.2. Marginal Tax Rates

In Section 2, we hypothesized that firms with limited ability to use tax shields are those for which the benefits of lease financing is highest. Table 5 examines the relationship between marginal tax rates and the impact of leasing and debt on borrowing costs. We obtain marginal tax rates before financing from John Graham. In Panel A of Table 5, we create orthogonal variables (high tax rate firms and low tax rate firms) for TD/TA and Leases/TA based on their tax status. Specifically, TD/TA for high tax rate firms is equal to TD/TA if the firm’s tax rate is above or equal to the median and 0 otherwise. TD/TA for low tax rate firms is equal to TD/TA if the firm’s tax rate is below the median and 0 otherwise. Similarly, Leases/TA for high tax rate 17

ACCEPTED MANUSCRIPT firms is equal to Leases/TA if the firm’s tax rate is above or equal to the median, 0 otherwise; Leases/TA for low tax rate firms is equal to Leases/TA if the firm’s tax rate is below the median

PT

and 0 otherwise. We observe that for low tax rate firms, the coefficient on debt is statistically

RI

and economically higher than the coefficient on leases, 2.49 vs. 1.63 with p-value of .091 using the S&P method, while the coefficients of debt and leases for the high marginal tax rate firms are

SC

essentially the same (1.80 vs. 1.86 with a p-value of .84). The results are confirmed using the

NU

Moody’s and the LMM methods. The results are broadly consistent with the idea that leases are more beneficial for low marginal tax rate firms vis-à-vis debt financing compared to high

MA

marginal tax rate firms. We note that the median marginal tax rate is 35% and it is also the mode and comprises the majority of the observations between the 25th and the 75th percentile. In order

D

to eliminate the lack of variability in marginal tax rates, we also perform two different

TE

specifications: in Panel B of Table 5, “high tax rate” firm are those with firm’s tax rate above the

AC CE P

67th percentile (excluding those observations equal to 35%) while “low tax rate” firms are those below the 33rd percentile; in Panel C of Table 5 “high tax rate” firms are those with tax rate greater than 35%, and “low tax rate” firms are those with tax rate less than 35%. Those two panels confirm the inferences from Panel A that leases are less sensitive to the cost of new borrowing than debt is in the case of low marginal tax rate firms. In unreported regressions, we used the abnormal component of leasing instead of the actual amount. Overall, the results are robust to this specification of leasing as the difference between the effects of debt vis-à-vis leasing on the cost of new borrowing is greater for low tax rate firms than for high tax rate firms.

18

ACCEPTED MANUSCRIPT 6. Debt, leases, and credit ratings In this section, we examine the effect of debt and leases on credit risk, as assessed by

PT

credit ratings. Subsection 6.1 examines the effects of debt and lease financing on existing debt

RI

ratings as well as new issue ratings and subsection 6.2 examines the propensity to lease as a

SC

function of credit ratings, especially for those firms that are on the boundary of rating categories.

NU

6.1. The Effect of Leasing and Debt on Credit Ratings

We test the hypothesis developed in Section 2 about the effect of debt and leases on

MA

credit ratings. We expect that all else equal, credit risk, and therefore, credit ratings, will drop with increased leverage, whether on-balance sheet debt financing or off-balance sheet lease

TE

bond ratings.

D

financing. Table 6 presents test results for existing ratings and Table 7 presents results for new

AC CE P

We obtain S&P credit ratings from COMPUSTAT and assign them numerical ranking in ascending order starting with 1 for the lowest S&P rating and ending at 26 for the highest S&P rating (i.e., AAA). In addition to debt and leases, we control for other factors affecting ratings such as firm total assets, profitability, and asset tangibility (measured by Net PPE/TA). Table 6 presents both OLS and ordered logit results, as the credit categories actually represent a discrete variable. We find that larger firms, firms with more tangible assets, and more profitable firms have higher ratings. We find that both debt and leases affect existing ratings, but debt has a higher impact on credit risk than operating leases. Table 7 reports results from a similar analysis for the ratings on newly issued bonds. We obtain the S&P and Moody ratings for the SDC bond issues made by our COMPUSTAT sample, and assign them numerical ranking in ascending order starting with 1 for the lowest S&P rating and ending at 26 for the highest S&P rating (i.e., AAA). In Table 7, we present the results from 19

ACCEPTED MANUSCRIPT OLS analysis. In unreported analysis, we obtain similar results from ordered logit regressions. We confirm our prior results for existing ratings: larger firms, firms with more tangible assets,

PT

and more profitable firms have higher ratings. We find that both debt and leases affect ratings

RI

for newly issued bonds, but once again debt has higher impact on credit risk than operating leases. As with the evidence from borrowing costs, we find that both debt and leases affect

SC

ratings, but leases affect credit ratings less than debt. We repeated the analyses in Tables 6 and 7

NU

with the abnormal components of leasing. In all specifications the effect of the abnormal

MA

component of leasing on ratings was significantly lower than that of debt. 6.2. Firms with Borderline Ratings

D

Given the evidence that leases impact credit ratings less than debt, we examine whether

TE

firms on the border of credit rating categories tend to use more leases, either to improve ratings

AC CE P

or to prevent ratings downgrades. Graham and Harvey (2001) document that credit ratings and financial flexibility (which they interpret to mean preserving unused debt capacity) are the two most important factors affecting debt policy in a survey of U.S. firms. Kisgen (2006) finds that firms with a credit rating designated with a plus or minus (e.g., AA+ or AA–) issue less debt relative to equity than firms that do not have a plus or minus (e.g., AA). While Kisgen (2006) focused on the choice of debt or equity in the total capital structure, we instead focus on the choice of leasing or debt in the credit structure. We present these results in Table 8. We examine whether the ratio of operating leases to the sum of balance sheet debt and leases is affected by whether a firm has a rating on the border of a rating category, i.e., those with “+” or “–” modifiers in their ratings, while controlling for firm size, asset tangibility, and credit rating. These three variables are suggested by Rampini and Viswanathan (2013) as determinants of the propensity to lease. We conduct a Tobit analysis (as the ratio is censored 20

ACCEPTED MANUSCRIPT between 0 and 1) as well as OLS. We obtain similar results from both analyses, and for all three lease estimation methods. For brevity and ease of interpretation, we only report the OLS results

PT

for the S&P and Moody’s method.

RI

We observe that the dummy variable, set to 1 if the firm has a borderline rating and 0 otherwise, is positive and significant. On average, firms on the border have 0.8% more leases

SC

relative to (debt + leases) in Panel A of Table 8. This result is economically meaningful also,

NU

considering that the overall average ratio of leases/(debt + leases) is around 16% for all rated firms in the sample. This result is mainly driven by the firms on the “–” side of ratings: the

MA

dummy takes 1 if the rating is followed by a “–” and 0 otherwise. The slope on the “–” rating variable is highly significant and indicates that firms in the “–” category use 1.8% more leases in

TE

D

their total obligations.7 The investment grade border is a dummy variable for rating close to the investment grade borderline. The dummy is equal to 1 if the credit rating is BBB– (lowest

AC CE P

investment grade) or BB+ (highest speculative grade) and 0 otherwise. Finally, we focus on the extreme border – those firms around the investment grade rating, i.e., BBB– and BB+. The regression results in Table 8 indicate that the investment-grade borderline firms use more leasing than firms not on the investment-grade borderline. The OLS results indicate that BBB– firms use 4.4% (5.2%) more leasing, while BB+ firms use 2.8% (2.7%) more leasing, using S&P (Moody’s) capitalization. The economic significance of 2.7% to 5.2% more lease usage is of similar magnitude as the increased issuance Kisgen (2006) reports for plus or minus ratings firms. We interpret the results as suggesting that firms “close” to ratings borderlines use more leasing in an apparent attempt to help maintain their credit rating. We obtain similar evidence from the decomposition 7

We exclude the AAA category from the calculations of plus and minus categories as there are no modifiers for AAA category. The results are not changed if we include that category. Further, the category is not excluded in the tests involving the investment grade border.

21

ACCEPTED MANUSCRIPT of leases into predicted and abnormal components, but the statistical significance is sensitive to the scaling (the denominator). The evidence that borderline firms use leases to maintain credit

PT

ratings is consistent with recent studies on earnings management. For example, Jung et al.

ratings.

RI

(2013) report that bond issuers smooth reported earnings to improve or maintain their credit Similarly, Alissa et al. (2013) report that firms use income-increasing (income-

SC

decreasing) earnings management activities such as abnormal accruals or real activities earnings

NU

management when they are below (above) their expected credit ratings. They further find that such earnings management actions are successful in helping those firms move toward their

MA

expected credit ratings.

While firms can manage earnings to massage credit metrics, operating leases would seem

TE

D

to be different. Given that the credit ratings agencies explicitly incorporate operating leases in their credit ratings methodologies, it is somewhat of a puzzle that firms use more operating

AC CE P

leases when on ratings borderlines. One possible explanation is that the ratings agencies do not capitalize the full debt equivalent of the lease, as suggested by the classic accounting textbook by Resvine et al. (2014, p. 689):

Lessees prefer the operating lease method for lease accounting. One obvious reason is that the operating lease method doesn’t reflect the cumulative liability for all future lease payments on the lessee’s balance sheet. Instead, only a portion of the obligation gets accrued piecemeal as partial performance takes place under the lease.

Another potential explanation is that the ratings agencies treat their estimated leasing debt equivalent more favorably than balance sheet debt (consistent with the results in Tables 6 and 7), perhaps due to the increased financial flexibility. It will be interesting to see if ratings borderline firms continue to use more leasing once leasing is brought onto the balance sheet.

22

ACCEPTED MANUSCRIPT 7. Conclusion We investigate how credit markets incorporate information about operating leases based

PT

on the current accounting disclosure rules. We find that the market treats operating leases

RI

differently than debt, as reflected in the pricing of new corporate bonds and bank loans. We find that both balance sheet debt and off-balance sheet leasing are important determinants of

SC

borrowing costs in public bond issues and bank loans. The borrowing cost of new debt is

NU

positively correlated with existing borrowings, whether the borrowing is on- or off-balance sheet. However, balance sheet debt has a larger impact on borrowing costs than does leasing,

MA

particularly for firms that are financially constrained or have low marginal tax rates. The evidence from borrowing costs is consistent with theoretical predictions that a dollar of leases

capacity for some firms.

TE

D

displaces less than a dollar of debt, so that leasing is associated with expanded total credit

AC CE P

We also examine the impact of leasing on credit ratings, and find that, even though balance sheet debt has a larger impact on credit ratings than does lease financing. Although credit ratings agencies explicitly state that they incorporate operating leases into their ratings criteria, we find the impact of leases on credit ratings is still less than that of on-balance sheet debts. Lastly, we find evidence that firms closer to ratings borderlines are more likely to lease, particularly when their credit ratings are on the investment grade borderline. Our results suggest that the economic advantage of leases arises from the nature of the true leases, whether leases are reported on or off the balance sheet. This view is consistent with Caskey and Ozel (2015), who find that the primary motives for leasing in the airline industry are economic in nature, placing reporting incentives secondary.

23

ACCEPTED MANUSCRIPT The economic benefits that we document lead us to predict that firms will continue to use leasing as an important source of credit, independent of the accounting treatment. It will be an

PT

interesting question as to how the new accounting rules will implement the mandate to capitalize

RI

operating leases on the balance sheet, and how those rules may impact future firm usage of

SC

leasing.

NU

Acknowledgments

MA

The authors thank the Charles Tandy American Enterprise Center for financial support. Mann and Mihov thank the Beasley Foundation and the Luther King Capital Management Center for

References

TE

D

Financial Studies for financial support.

AC CE P

Alissa, W., S. B. Bonsall, K. K. Koharki, and M. W. Penn Jr., 2013. Firms’ use of accounting discretion to influence their credit ratings, Journal of Accounting and Economics 55, 129-147. Altamuro, J., R. Johnston, S. Pandit, and H. Zhang. 2014. Operating leases and credit assessment. Contemporary Accounting Research 31 (2), p. 551-580. Ang, J. and P. Peterson, 1984. The Leasing Puzzle. Journal of Finance 39, 1055-1065. Barry, C., S. Mann, V. Mihov and M. Rodriguez, 2008. Corporate Debt Issuance and the Historical Level of Interest Rates, Financial Management 37, 413-430. van Binsbergen, J. H., J. R. Graham and J. Yang, 2010, The Cost of Debt, Journal of Finance 65, p. 2089–2136. Bratten, B., P. Chowdhary, and K. Schipper, 2013. Evidence that market participants assess recognized and disclosed items similarly when reliability is not an issue. The Accounting Review 88 (4): 1179-1210. Caskey, J., and N. B. Ozel, 2015. Reporting and non-reporting incentives in leasing: Evidence from the airline industry, University of Texas working paper.

24

ACCEPTED MANUSCRIPT Cornaggia, K., L. Franzen, and T. Simin. 2013. Bringing leased assets onto the balance sheet. Journal of Corporate Finance 22: 345-360.

PT

Demerjian, P., 2011, Accounting standards and debt covenants: Has the “balance sheet approach” led to a decline in the use of balance sheet covenants? Journal of Accounting and Economics 52, 178-202.

RI

Denis, D, and J. Wang, 2014. Debt covenant renegotiations and creditor control rights, Journal of Financial Economics 113, 348-367.

SC

Financial Accounting Standards Board, 1976. Accounting for Leases. Statement of Financial Accounting Standards No. 13. Stamford, CT: FASB.

NU

Eisfeldt, A. and A. Rampini, 2009. Leasing, Ability to Repossess, and Debt Capacity. Review of Financial Studies 22, 1621-1657.

MA

Faulkender, M, and M. Petersen, 2006. Does the Source of Capital Affect Capital Structure?, Review of Financial Studies 19, 45-79.

TE

D

Graham, J., C. Harvey, 2001. The theory and practice of corporate finance: evidence from the field, Journal of Financial Economics 60, 187-243.

AC CE P

Graham, J., M. Lemmon, and J. Schallheim, 1998. Debt, Leases, and the Endogeneity of Corporate Tax Status, Journal of Finance 53 (1), 131-162. Imhoff, E. and J. K. Thomas, 1988. Economic consequences of accounting standards: The lease disclosure rule change, Journal of Accounting and Economics 10, 277-310. Jung, B., N. Soderstrom, and Y. S. Yang, 2013. Earnings smoothing activities of firms to manage credit ratings, Contemporary Accounting Research 30 (2), 645-676. Kisgen, D. J, 2006. Credit rating and capital structure, The Journal of Finance 61 (3), 10351072. Lewis, C., and J. Schallheim, 1992. Are Debt and Leases Substitutes?, Journal of Financial and Quantitative Analysis 27 (4), 497-511. Lim, S., S. Mann, and V. Mihov, 2003. Market evaluation of off-balance sheet financing: You can run but you can’t hide. Working paper, Texas Christian University. Marston, F. and R. Harris, 1988. Substitutability of leases and Debt in Corporate Capital Structures. Journal of Accounting, Auditing and Finance 3, 147-164. Miller, M. H., 1991. Leverage, Journal of Finance 46, 479-488.

25

ACCEPTED MANUSCRIPT Miller, M. and C. Upton. 1976. Leasing, Buying and the Cost of Capital Services. Journal of Finance 31, 761-786.

PT

Modigliani, F. and M. Miller, 1958. The cost of capital, corporation finance and the theory of investment, American Economic Review 48, 261-296.

RI

Myers, S., D. Dill and A. Bautista, 1976. Valuation of Financial Lease Contracts, Journal of Finance 31, 799-819.

SC

Plumlee, M., Y. Xie, M. Yan, and J. Yu. 2015. Bank Loan Spread and Private Information: Pending Approval Patents. Review of Accounting Studies 20: 593-638.

NU

Rajan, R. and A. Winton, 1995, Covenants and Collateral as Incentives to Monitor, Journal of Finance 50 (4), 1113-1146.

MA

Rampini, A., and S. Viswanathan, 2013. Collateral and Capital Structure, Journal of Financial Economics 109, 466-492.

D

Rauh, J., and A. Sufi. 2010. Explaining Corporate Capital Structure: Product Market, Leases, and Asset Similarity. Review of Finance 16, 115-155.

TE

Reisel, N, 2014. On the value of restrictive covenants: Empirical investigation of public bond issues, Journal of Corporate Finance 27, 251-268. Financial

AC CE P

Resvine, L., D. Collins, W.B. Johnson, H.F. Mittelstaedt, and L.C.Sofer, 2014. Reporting and Analysis, 6th Edition, McGraw-Hill, New York.

Schallheim, J, K. Wells, and R. Whitby, 2013. Do Leases Expand Debt Capacity? Journal of Corporate Finance 23, 368-381. Smith, C. and L. Wakeman, 1985. Determinants of Corporate Leasing Policy. Journal of Finance 40, 895-908. Yan, A., 2006. Leasing and Debt Financing: Substitutes or Complements? Journal of Financial and Quantitative Analysis 41, 709-731. Zechman, S. 2010. The relation between voluntary disclosure and financial reporting: evidence from synthetic leases. Journal of Accounting Research 48: 725-765.

26

D

MA

NU

SC

RI

PT

ACCEPTED MANUSCRIPT

AC CE P

TE

Figure 1. Cost of safe, unsecured, and secured credit.

27

ACCEPTED MANUSCRIPT

AC

CE

PT ED

MA

NU

SC

RI

PT

Table 1 Descriptive Statistics. Descriptive statistics for a sample of 31,339 COMPUSTAT firm/years in SIC codes 2000-5999, during 1995-2011. All dollar values are measured in constant 2011 dollars. Total debt is equal to short-term debt plus long-term debt. The S&P value of operating leases are measured as the present value of the future minimum obligations discounted at yield on a five year bond with the same rating as the firm. For the unrated firms, we estimate a rating and use the rating-matched yield. The Moody’s value of operating lease is the higher value between S&P’s and 8 times current year rent expense. The LMM value is based on a depreciation-adjusted perpetuity. TA does not include the debt equivalent value of operating leases. Investment SubDid Not Issued No Dealscan Means Full Sample Grade Investment All Rated Not Rated Issue SDC SDC Dealscan Loans Rating Grade Rating bonds bonds Loans N 31,339 5,968 6,733 12,701 18,638 23,299 8,040 10,071 21,328 Means Sales 7,987.2 30,477.8 4,214.4 17,568.1 1,677.1 4,484.8 17,421.3 4,653.8 9,494.7 -0.035 0.121 EBITDA/TA 0.071 0.155 0.112 0.132 0.029 0.047 0.138 0.305 0.321 Total Debt/TA 0.316 0.260 0.493 0.384 0.270 0.305 0.350 0.097 0.099 Leases (S&P method)/TA 0.099 0.061 0.096 0.080 0.112 0.105 0.079 0.274 0.240 Leases (Moody’s method)/TA 0.251 0.140 0.237 0.191 0.292 0.277 0.176 Leases (LMM method)/TA 0.212 0.155 0.199 0.177 0.235 0.227 0.169 0.215 0.210 Medians Sales ($ mil) 987.2 10,872.1 1,807.9 4,322.5 308.1 478.9 4,905.6 189.8 1,613.2 0.068 0.125 EBITDA/TA 0.113 0.148 0.111 0.128 0.097 0.103 0.134 0.219 0.275 Total Debt/TA 0.262 0.251 0.446 0.329 0.204 0.240 0.308 0.038 0.036 Leases (S&P method)/TA 0.036 0.028 0.036 0.031 0.042 0.039 0.031 0.134 0.113 Leases (Moody’s method)/TA 0.119 0.083 0.112 0.095 0.143 0.132 0.095 Leases (LMM method)/TA 0.101 0.089 0.092 0.091 0.111 0.106 0.093 0.102 0.101

28

ACCEPTED MANUSCRIPT

T-value 36.04 -36.50 -32.02 4.94 -13.87 3.60 13.26 -14.59 -1.59 25.24 -22.40 31.42

St. Coeff. 0.000 -0.320 -0.234 0.036 -0.112 0.027 0.095 -0.112 -0.013 0.204 -0.160 0.237

25.077

6.48

0.048

NU

Coeff. 827.310 -28.014 -398.617 28.522 -94.782 5.699 0.445 -40.293 -4.229 77.285 -467.981 147.129

MA

PT ED

St. Coeff. 0.000 -0.326 -0.233 0.036 -0.112 0.026 0.094 -0.112 -0.015 0.206 -0.160 0.237 0.037

CE

T-value 36.41 -37.66 -31.85 4.81 -13.81 3.54 13.11 -14.56 -1.76 25.47 -22.40 31.30 5.03

AC

Intercept Log(TA) EBITDA/TA Net PPE/ TA Amount Issued/ TA Log(Maturity) TED Spread Balance sheet covenants dummy Earnings covenants dummy Investment covenants dummy Seniority dummy Total Debt / TA Leases (S&P method)/TA Leases (Moody’s method)/TA Leases (LMM method)/TA Number of observations Adj. R-square

Coeff. 834.197 -28.539 -396.408 28.066 -94.423 5.606 0.441 -40.249 -4.691 77.982 -468.426 146.936 38.029

SC

RI

PT

Table 2 The Effect of Existing Debt and Leases on the Cost of New Bank Loans. The dependent variable is the “All-in-Drawn Spread,” in basis points, for loan facilities reported by LPC Dealscan database for a sample of US firms in SIC codes 2000-5999 with available COMPUSTAT data. All dollar values are in constant 2011 dollars. Total debt is equal to short-term debt plus long-term debt. The S&P value of operating leases is the present value of the future minimum lease obligations discounted at the yield on a five year bond with the same rating as the firm. For unrated firms, we estimate a rating and use rating-matched yields. The Moody’s value of operating lease is estimated using a multiple. The LMM method is a depreciation adjusted perpetuity. Loan maturity is measured in months. The TED Spread is the spread between 3-Month LIBOR based on US dollars and 3-Month Treasury Bill, as reported by St. Louis FRED database. The covenants dummies are created following the classification by Damerjan (2011). The seniority dummy is equal to 1 if the loan is senior, 0 otherwise. In Panel B, we decompose each lease measure into a predicted and “abnormal” component, following Cornaggia, Franzen, and Simin (2013), and use the abnormal component in the yield regression. Panel A S&P Moody’s LMM

12,459 0.369

12,459 0.370

Coeff. 851.023 -28.374 -394.382 29.372 -90.600 6.948 0.470 -37.932 -4.215 75.986 -495.408 147.778

T-value 36.75 -36.45 -30.40 4.89 -13.04 4.30 13.72 -13.60 -1.56 24.43 -23.51 29.99

St. Coeff. 0.000 -0.328 -0.231 0.038 -0.109 0.033 0.102 -0.109 -0.013 0.204 -0.174 0.236

12.274 11,600 0.370

2.59

0.020

Test for equality of the coefficients of Total Debt/ TA and Leases/ TA: F-value (p-value)

F-value (p-value)

F-value (p-value)

165.55 (0.000)

442.41 (0.000)

452.23 (0.000)

29

ACCEPTED MANUSCRIPT

5,950 0.390

St. Coeff. 0.000 -0.348 -0.207 0.053 -0.113 0.054 0.106 -0.119 0.003 0.246 -0.072 0.238

14.704

2.38

0.025

SC

RI

PT

Coeff. 618.861 -29.250 -345.007 41.541 -95.618 10.655 0.452 -37.637 0.774 89.983 -307.982 167.886

Moody’s T-value 13.58 -28.28 -19.69 5.05 -9.81 5.06 10.37 -10.82 0.22 21.45 -7.07 22.10

NU

St. Coeff. 0.000 -0.347 -0.206 0.054 -0.113 0.053 0.106 -0.118 0.002 0.246 -0.072 0.238 0.029

MA

Coeff. 622.963 -29.190 -344.070 41.985 -95.458 10.536 0.451 -37.409 0.628 90.115 -309.836 167.832 33.249

CE

Intercept Log(TA) EBITDA/TA Net PPE/ TA Amount Issued/ TA Log(Maturity) TED Spread Balance sheet covenants dummy Earnings covenants dummy Investment covenants dummy Seniority dummy Total Debt / TA “Abnormal” Leases (S&P method)/TA “Abnormal” Leases (Moody’s method)/TA “Abnormal” Leases (LMM method)/TA Number of observations Adj. R-square

S&P T-value 13.75 -28.20 -19.69 5.11 -9.79 5.00 10.36 -10.75 0.18 21.49 -7.12 22.15 2.79

PT ED

Panel B

Coeff. 638.094 -29.349 -336.543 39.741 -95.367 10.670 0.459 -38.158 0.428 89.565 -309.907 159.850

LMM T-value 14.06 -28.10 -19.01 4.75 -9.70 4.97 10.37 -10.86 0.12 21.03 -7.14 20.26

St. Coeff. 0.000 -0.353 -0.204 0.051 -0.114 0.054 0.109 -0.122 0.001 0.246 -0.074 0.223

-0.79

-0.008

-5.717 5,780 0.381

5,950 0.389

Test for equality of the coefficients of Total Debt/ TA and “Abnormal” Leases/ TA:

100.27 (0.000)

AC

F-value (p-value)

F-value (p-value)

F-value (p-value)

278.16 (0.000)

279.10 (0.000)

30

ACCEPTED MANUSCRIPT

9.67 -10.14 -20.80 -0.35 7.85 -4.06 13.12 25.85 14.19 9.02

0.000 -0.179 -0.263 -0.004 0.104 -0.050 0.236 0.339 0.191 0.114

3,399 0.485

CE

T-value

St. Coeff.

Coeff.

T-value

St. Coeff.

2.292 -0.222 -7.041 -0.027 0.441 -0.773 4.544 0.566 2.050

8.94 -9.53 -20.93 -0.25 8.00 -4.08 13.06 25.88 14.26

0.000 -0.168 -0.263 -0.003 0.106 -0.050 0.234 0.337 0.191

2.328 -0.228 -6.824 -0.079 0.436 -0.748 4.630 0.569 2.107

8.70 -9.31 -18.96 -0.69 7.59 -3.88 12.75 25.16 14.01

0.000 -0.172 -0.249 -0.009 0.105 -0.050 0.237 0.344 0.195

1.228

10.65

0.134 1.032 3,215 0.470

7.93

0.107

NU

2.474 -0.236 -7.029 -0.038 0.434 -0.773 4.588 0.568 2.051 1.804

Coeff.

MA

St. Coeff.

PT ED

T-value

AC

Intercept Log(TA) EBITDA/TA Net PPE / TA Callable dummy Puttable dummy Amount Issued/ TA BAA minus GS10 Total Debt / TA Leases (S&P method)/TA Leases (Moody’s method)/TA Leases (LMM method)/TA Number of observations Adj. R-square

Coeff.

SC

RI

PT

Table 3 The Effect of Existing Debt and Leases on Borrowing Cost of New Bonds. The dependent variable is the yield to maturity on newly issued bonds in excess of the yield of a maturity-matched Treasury for a sample of US firms in SIC codes 2000-5999 that issued public debt during 1995-2011 as reported by Thompson Financial SDC database with available COMPUSTAT data. All dollar values are measured in constant 2011 dollars. Total debt is equal to short-term debt plus long-term debt. The S&P value of operating leases is the present value of the future minimum lease obligations discounted at the yield on a five year bond with the same rating as the firm. For unrated firms, we estimate a rating and use rating-matched yields. The Moody’s value of operating lease is estimated using a multiple. The LMM method is a depreciation adjusted perpetuity. We include dummy variables for callability and puttability of the new issues. BAA minus GS10 is the spread between the composite Moody's Seasoned Baa Corporate Bond and the 10-year constant maturity Treasury, as reported by St. Louis FRED database. In Panel B, we decompose each lease measure into a predicted and “abnormal” component, following Cornaggia, Franzen, and Simin (2013), and use the abnormal component in the yield regression. Panel A S&P Moody’s LMM

3,399 0.490

Test for equality of the coefficients of Total Debt/ TA and Leases/ TA: F-value (p-value)

F-value (p-value)

F-value (p-value)

0.95 (0.329)

19.11 (0.000)

28.92 (0.000)

31

ACCEPTED MANUSCRIPT

S&P

PT

1.881 -0.235 -6.831 0.163 0.477 -0.712 4.248 0.573 2.198

6.17 -8.83 -17.82 1.29 7.90 -3.69 9.14 23.75 12.83

0.000 -0.179 -0.265 0.019 0.126 -0.054 0.190 0.371 0.199

5.38

0.078

RI

0.000 -0.174 -0.262 0.022 0.123 -0.054 0.192 0.372 0.195 0.081

LMM

T-value St. Coeff.

SC

6.67 -8.57 -17.67 1.47 7.71 -3.67 9.23 23.80 12.57 5.50

Coeff.

NU

Intercept 1.987 Log(TA) -0.229 EBITDA/TA -6.769 Net PPE / TA 0.186 Callable dummy 0.465 Puttable dummy -0.708 Amount Issued/ TA 4.293 BAA minus GS10 0.574 Total Debt / TA 2.153 “Abnormal” Leases (S&P method)/TA 1.296 “Abnormal” Leases (Moody’s method)/TA “Abnormal” Leases (LMM method)/TA Number of observations 2,627 Adj. R-square 0.450

T-value St. Coeff.

MA

Coeff.

Moody’s

PT ED

Panel B

0.651 2,627 0.450

Coeff.

T-value St. Coeff.

2.021 -0.232 -6.597 0.156 0.486 -0.710 4.066 0.579 2.188

6.53 -8.48 -16.26 1.21 7.78 -3.63 8.46 23.48 12.47

0.000 -0.177 -0.249 0.019 0.129 -0.055 0.181 0.379 0.199

0.509 2,508 0.438

3.61

0.054

Test for equality of the coefficients of Total Debt/ TA and “Abnormal” Leases/ TA: F-value (p-value)

F-value (p-value)

55.27 (0.000)

56.75 (0.000)

AC

8.42 (0.004)

CE

F-value (p-value)

32

ACCEPTED MANUSCRIPT

NU

SC

RI

PT

Table 4 Cost of Debt and Proxies for Financial Constraints. The dependent variable is the yield spread of new bonds over a maturity-matched Treasury for firms that issued public debt during 1995-2011 as reported by Thompson Financial. In Panel A, we classify firms based on their free cash flow (FCF). TD/TA for high FCF firms is equal to TD/TA if the firm had a ratio of FCF to total assets above or equal to the median, 0 otherwise; TD/TA for low FCF firms is equal to TD/TA if the firm had a ratio of FCF to total assets below the median, 0 otherwise; Leases/TA for high FCF firms is equal to Leases/TA if the firm had a ratio of FCF to total assets above or equal to the median, 0 otherwise; Leases/TA for low FCF firms is equal to Leases/TA if the firm had a ratio of FCF to total assets below the median, 0 otherwise. In Panel B, we classify the firm based on existing rating. TD/TA for rated firms is equal to TD/TA if the firm had rated debt, 0 otherwise; TD/TA for unrated firms is equal to TD/TA if the firm did not have rated debt, 0 otherwise; Leases/TA for rated firms is equal to Leases/TA if the firm had rated debt, 0 otherwise; Leases/TA for unrated firms is equal to Leases/TA if the firm did not have rated debt, 0 otherwise; In Panel C, we classify firms based on their interest service. “Low interest” is a dummy variable equal to 1 if the firm has a ratio of interest to total assets below the sample median, 0 otherwise. TD/TA for high interest firms is equal to TD/TA if the firm had a ratio of interest to total assets greater than or equal to the sample median, 0 otherwise; TD/TA for low interest firms is equal to TD/TA if the firm had a ratio of interest to total assets lower than the sample median, 0 otherwise; Leases/TA for high interest firms is equal to Leases/TA if the firm had a ratio of interest to total assets greater than or equal to the sample median, 0 otherwise; Leases/TA for low interest firms is equal to Leases/TA if the firm had a ratio of interest to total assets lower than the sample median, 0 otherwise.

MA

S&P

Panel A: Free Cash Flow

t-value

Estimate

t-value

Estimate

t-value

2.427

9.49

2.239

8.74

2.311

8.65

1.533

9.25

1.539

9.27

1.565

8.97

2.089

7.20

1.268

8.34

1.052

6.13

2.543

15.48

2.527

15.33

2.589

15.06

1.420

5.33

1.131

6.78

0.943

5.20

0.570

25.94

0.567

25.96

0.570

25.22

-10.24

-0.224

-9.62

-0.234

-9.57

-6.051

-16.11

-6.008

-16.06

-5.727

-14.29

-0.199

-1.79

-0.202

-1.82

-0.253

-2.13

0.454

8.19

0.463

8.39

0.463

8.05

-0.826

-4.35

-0.821

-4.35

-0.798

-4.16

4.479

12.81

4.436

12.76

4.458

12.26

D

TD/TA, high FCF firms

Leases/TA, low FCF firms

AC CE P

BAAminusGS10 Log(TA)

TE

TD/TA, low FCF firms

LMM

Estimate

Intercept Leases/TA, high FCF firms

Moody’s

-0.238

F test (TD/TA=Leases/TA, high FCF firms), p-value in parentheses

2.49

(0.115)

1.26

(0.262)

3.88

(0.049)

10.87

(0.001)

28.89

(0.000)

35.93

(0.000)

EBITDA/TA Net PPE/TA Callable Puttable Issue size N Adj. R-square

F test (TD/TA=Leases/TA, low FCF firms), pvalue in parentheses

3,378

3,378

3,194

0.491

0.498

0.477

33

ACCEPTED MANUSCRIPT S&P Estimate

t-value

Estimate

LMM

t-value

Estimate

t-value

2.433

9.54

2.247

8.80

2.316

8.69

TD/TA, rated firms

1.953

13.35

1.928

13.20

1.993

13.00

Leases/TA, rated firms

1.830

8.84

1.285

10.53

1.033

7.62

TD/TA, unrated firms

3.163

11.56

3.255

11.89

3.239

11.43

Leases/TA, unrated FCF firms

1.678

2.44

0.833

2.66

1.093

2.78

BAAminusGS10 Log(TA)

0.575

26.21

0.572

26.24

0.575

25.47

-0.232

-9.98

-0.217

-9.35

-0.225

-9.26

EBITDA/TA

-6.909

-20.49

-6.916

-20.61

-6.698

-18.63

Net PPE/TA

-0.43

-0.039

-0.36

-0.080

-0.69

0.430

7.80

0.435

7.93

0.435

7.61

Puttable

-0.765

-4.04

-0.765

-4.06

-0.744

-3.88

Issue size

4.184

11.65

4.198

11.69

4.153

11.02

N

3,397

MA

-0.047

Callable

NU

SC

PT

Intercept

RI

Panel B: Existing Rated Debt

Moody’s

0.489

F test (TD/TA=Leases/TA, rated firms), pvalue in parentheses

AC CE P

TE

F test (TD/TA=Leases/TA, unrated firms), pvalue in parentheses

D

Adj. R-square

Panel C: Interest Service

3,397

3,215

0.494

0.475

0.22

(0.638)

10.52

(0.001)

20.91

(0.000)

3.34

(0.068)

26.11

(0.000)

15.35

(0.000)

S&P

Moody’s

Estimate

t-value

Estimate

2.502

9.77

2.299

TD/TA, low interest obligation firms

0.934

3.24

0.942

Leases/TA, low interest obligation firms

1.969

6.01

TD/TA, high interest obligation firms

1.865

11.87

Leases/TA, high interest obligation firms

1.650

BAAminusGS10 Log(TA)

0.573

Intercept

LMM

t-value

Estimate

t-value

8.95

2.340

8.73

3.26

1.011

3.34

1.526

7.35

1.158

5.67

1.937

12.24

1.948

11.81

6.60

1.057

7.82

0.925

5.77

25.96

0.571

25.98

0.574

25.24

-0.224

-9.55

-0.209

-8.90

-0.214

-8.69

EBITDA/TA

-6.953

-20.39

-7.000

-20.62

-6.761

-18.56

Net PPE/TA

-0.077

-0.70

-0.067

-0.63

-0.121

-1.04

Callable

0.448

8.07

0.451

8.16

0.444

7.70

Puttable

-0.805

-4.23

-0.799

-4.22

-0.778

-4.04

Issue size

4.630

13.19

4.572

13.11

4.687

12.86

N

3,375

3,375

3,192

Adj. R-square

0.488

0.492

0.473

F test (TD/TA=Leases/TA, low interest firms), p-value in parentheses

5.31

(0.021)

2.50

(0.114)

0.15

(0.703)

F test (TD/TA=Leases/TA, high interest firms), p-value in parentheses

0.46

(0.499)

15.32

(0.000)

16.92

(0.000)

34

ACCEPTED MANUSCRIPT

S&P

Moody’s

t-value

Estimate

SC

Estimate

LMM

t-value

Estimate

t-value

2.135

7.78

1.955

7.12

2.025

7.07

TD/TA, high tax rate firms

1.795

10.29

1.786

10.20

1.770

9.71

Leases/TA, high tax rate firms

1.857

7.89

1.258

9.58

1.133

7.58

TD/TA, low tax rate firms

2.487

12.68

2.493

12.49

2.594

12.73

Leases/TA, low tax rate firms

1.626

3.92

1.135

4.36

0.767

2.84

BAAminusGS10 Log(TA)

0.607

26.14

0.604

26.19

0.609

25.57

-0.208

-8.33

-0.195

-7.79

-0.203

-7.73

-6.824

-18.31

-6.823

-18.38

-6.643

-16.78

-0.085

-0.72

-0.071

-0.61

-0.096

-0.78

0.406

6.97

0.415

7.17

0.407

6.76

-0.722

-3.76

-0.726

-3.80

-0.703

-3.62

4.384

10.33

4.277

10.14

4.331

9.85

MA

Intercept

NU

Panel A: Top 1/2 vs. Bottom 1/2

RI

PT

Table 5 Cost of Debt and Marginal Tax Rates. The dependent variable is the yield spread of new bonds over a maturity-matched Treasury for firms that issued public debt during 1995-20011 as reported by Thompson Financial. We obtain firm marginal tax rates before financing from John Graham. TD/TA for high tax rate firms is equal to TD/TA if the firm’s tax rate is above or equal to the median and 0 otherwise. TD/TA for low tax rate firms is equal to TD/TA if the firm’s tax rate is below the median and 0 otherwise. Leases/TA for high FCF firms is equal to Leases/TA if the firm had a ratio of FCF to total assets above the median, 0 otherwise; Leases/TA for low FCF firms is equal to Leases/TA if the firm had a ratio of FCF to total assets below the median, 0 otherwise. In Panel B, “high tax rate ” includes firms with tax rate above the 67 th percentile, and “low tax rate” below the 33rd percentile; in Panel C, “high tax rate ” includes firms with a tax rate is greater than 35%, and “low tax rate” less than 35%.

EBITDA/TA Net PPE/TA

D

Callable

N Adj. R-square

AC CE P

Issue size

TE

Puttable

2,959

2,959

2,818

0.473

0.479

0.462

F test (TD/TA=Leases/TA, high tax rate firms), p-value in parentheses

0.04

(0.842)

5.09

(0.024)

6.50

(0.011)

F test (TD/TA=Leases/TA, low tax rate firms), p-value in parentheses

2.87

(0.091)

12.85

(0.000)

22.41

(0.000)

35

ACCEPTED MANUSCRIPT S&P

Moody’s

LMM

Panel B: Top 1/3 vs. Bottom 1/3 3.557

7.98

3.343

7.55

3.274

7.10

TD/TA, high tax rate firms

2.335

8.37

2.306

8.32

2.183

7.61

Leases/TA, high tax rate firms

2.670

7.08

1.710

8.69

1.763

7.56

TD/TA, low tax rate firms

3.187

11.38

3.232

11.51

3.348

11.73

Leases/TA, low tax rate firms

2.105

3.59

1.344

3.90

1.048

2.88

BAAminusGS10 Log(TA)

0.689

19.15

0.688

19.28

0.707

19.33

-0.359

-8.93

-0.342

-8.55

-0.348

-8.36

EBITDA/TA

-8.344

-12.67

-8.396

-12.86

-7.596

-11.01

Net PPE/TA

-0.086

-0.46

-0.085

-0.43

RI

PT

Intercept

-0.41

0.310

3.25

0.314

3.32

0.292

3.00

Puttable

-0.594

-1.90

-0.599

-1.94

-0.541

-1.73

Issue size

2.588

3.99

2.636

4.03

N

1,436

Adj. R-square

0.512

NU

0.42

(0.519)

2.53

(0.112)

1.05

(0.306)

2.29

(0.130)

13.95

(0.000)

19.53

(0.000)

MA

TE

Panel C: Rate >35% vs. <35%

2.509

4.07

D

F test (TD/TA=Leases/TA, high tax rate firms), p-value in parentheses F test (TD/TA=Leases/TA, low tax rate firms), p-value in parentheses

SC

-0.078

Callable

S&P

1,436

1,387

0.520

0.506

Moody’s

LMM

Estimate

t-value

Estimate

t-value

Estimate

t-value

3.310

8.57

3.129

8.15

3.082

7.76

TD/TA, high tax rate firms

2.296

9.25

2.266

9.12

2.154

8.40

Leases/TA, high tax rate firms

2.699

7.55

1.729

9.26

1.764

8.03

TD/TA, low tax rate firms

2.687

11.58

2.714

11.68

2.816

11.95

Leases/TA, low tax rate firms

1.537

3.44

1.126

4.01

0.783

2.71

BAAminusGS10 Log(TA) EBITDA/TA Net PPE/TA

AC CE P

Intercept

0.691

21.33

0.689

21.47

0.707

21.47

-0.337

-9.54

-0.323

-9.20

-0.331

-9.10

-8.762

-15.60

-8.782

-15.78

-8.115

-13.91

-0.031

-0.19

-0.053

-0.33

-0.039

-0.23

Callable

0.412

5.15

0.415

5.24

0.402

4.96

Puttable

-0.748

-3.23

-0.748

-3.26

-0.710

-3.07

Issue size

3.017

5.29

2.922

5.18

3.036

5.19

N

1,727

1,727

1,677

Adj. R-square

0.524

0.532

0.518

F test (TD/TA=Leases/TA, high tax rate firms), p-value in parentheses

0.69

(0.408)

2.38

(0.123)

1.05

(0.305)

F test (TD/TA=Leases/TA, low tax rate firms), p-value in parentheses

4.36

(0.037)

14.97

(0.000)

23.92

(0.000)

36

ACCEPTED MANUSCRIPT

RI

PT

Table 6. The Effect of Debt and Leases on Existing Credit Ratings. The dependent variable is the existing S&P credit rating as reported by COMPUSTAT for a sample of US firms in SIC codes 2000-5999 during 1995-2011. Total debt is equal to short-term debt plus long-term debt. The S&P value of operating leases is the present value of the future minimum lease obligations discounted at the yield on a five year bond with the same rating as the firm. For unrated firms, we estimate a rating and use rating-matched yields. The Moody’s value of operating lease is estimated using a multiple. The LMM method is a depreciation adjusted perpetuity. We assign the S&P ratings a numerical ranking in descending order starting at 26. LMM method

St. Estimate

6.34

14.34

0.00

SC

Moody’s method

Log(TA)

1.19

80.34

0.49

1.17

77.6

0.48

1.17

76.25

0.49

EBITDA/TA

15.67

61.83

0.34

15.76

61.97

0.34

16.31

59.46

0.35

Net PPE/TA

0.36

3.61

0.02

0.22

2.22

0.01

0.34

3.25

0.02

Total Debt/TA

-4.63

-53.57

-0.32

MA

S&P method

-4.6

-53.12

-0.32

-4.72

-51.04

-0.32

-3.14

-21.3

-0.12

-1.47

-20.49

-0.12

-1.54

-18.46

-0.11

Leases/TA N R-square

t-value

St. Estimate

Estimate 6.56

12,507 0.63

t-value

Estimate

t-value

44.82

0.00

6.46

43.07

0.00

NU

Intercept

Estimate

PT ED

OLS

12,507

11,713

0.63

0.61

St. Estimate

F-test for equality of TD/TA and Leases/TA (p-values in parentheses) (0.000)

820.5

(0.000) p-value

735.3

(0.000)

Estimate

p-value

CE

82.2

Estimate

0.79

0.94

0.00

0.77

0.95

0.00

0.77

0.00

0.64

14.98

0.00

0.65

15.56

0.00

0.64

0.33

0.00

0.04

0.22

0.01

0.02

0.37

0.00

0.04

Total Debt/TA

-4.35

0.00

-0.6

-4.35

0.00

-0.60

-4.47

0.00

-0.59

Leases/TA

-2.74

0.00

-0.21

-1.41

0.00

-0.22

-1.50

0.00

-0.21

Ordered Logits

Estimate

p-value

0.96

0.00

EBITDA/TA

14.78

Net PPE / TA

N

St. Estimate

AC

Log(TA)

12,507

12,507

St. Estimate

St. Estimate

11,713

Chi-Square-test for equality of TD/TA and Leases/TA (p-values in parentheses) 146.2

(0.000)

1,040.2 (0.000)

37

882.5

(0.000)

ACCEPTED MANUSCRIPT

t-value

St. Estimate

Estimate

Estimate.

t-value

St. Estimate

Intercept

5.79

17.40

0.00

6.08

0.00

6.25

18.16

0.00

Log(TA)

1.23

-23.82

0.48

1.21

41.63

0.47

1.20

40.06

0.49

EBITDA/TA

21.18

36.14

0.39

Net PPE/TA

1.09

5.67

0.06

21.16

36.28

0.39

19.92

32.53

0.38

1.13

5.96

0.06

1.00

4.99

0.06

Total Debt/TA

-5.87

-23.82

-0.27

-5.86

-23.87

-0.27

-5.62

-22.11

-0.27

Leases/TA

-1.54

-4.23

-0.05

-1.45

-6.82

-0.07

-1.16

-5.09

-0.06

N

3,660

Adj. R-square

0.59

18.11

MA

NU

S&P Rating

t-value

St. Estimate

SC

Estimate.

RI

PT

Table 7 The Effect of Debt and Leases on New Credit Ratings. The dependent variable is the credit rating for new bond issues during 1995-2011 made by US firms in SIC codes 2000-5999 as reported by Thompson Financial, with available COMPUSTAT data. Total debt is equal to short-term debt plus long-term debt. The S&P value of operating leases is the present value of the future minimum lease obligations discounted at the yield on a five year bond with the same rating as the firm. For unrated firms, we estimate a rating and use rating-matched yields. The Moody’s value of operating lease is estimated using a multiple. The LMM method is a depreciation adjusted perpetuity. We assign the S&P and the Moody’s ratings a numerical ranking in descending order starting at 26. S&P method Moody’s method LMM method

3,449

0.59

0.55

PT ED

3,660

F-test for equality of TD/TA and Leases/TA (p-values in parentheses) 183.7

(0.000)

Moody’s Rating Intercept

5.13

16.35

Log(TA)

1.25

45.60

EBITDA/TA

21.74

Net PPE/TA

173.5

(0.000)

0.00

5.48

17.36

0.49

1.23

44.44

0.00

5.77

17.82

0.00

0.48

1.21

42.44

0.49

39.39

0.41

21.74

39.62

0.41

20.49

35.66

0.39

1.21

6.75

0.07

1.24

7.03

0.07

1.14

6.15

0.07

Total Debt/TA

-5.49

-23.36

Lease (S&P)/TA

-2.39

-7.19

-0.25

-5.48

-23.44

-0.25

-5.37

-22.26

-0.26

-0.08

-1.96

-9.84

-0.10

-1.65

-7.74

-0.09

N

3,965

3,965

3,752

Adj. R-square

0.59

0.59

0.55

CE

(0.000)

AC

96.8

F-test for equality of TD/TA and Leases/TA (p-values in parentheses) 56.1

(0.000)

126.8

38

(0.000)

134.2

(0.000)

ACCEPTED MANUSCRIPT

Est.

0.121 0.076 -0.011 0.006 0.008

11.36 8.97 -7.07 10.22 2.33

0.121 0.077 -0.011 0.007

MA

PT ED

-0.002 0.018

12,397 0.014 0.284 0.025 -0.025 0.013 0.004

23.16 2.61 -14.07 17.27 1.01

t-value

Est.

t-value

Est.

t-value

11.32 9.09 -7.23 10.42

0.129 0.080 -0.011 0.006

12.98 9.50 -7.49 9.98

0.130 0.080 -0.011 0.006

13.03 9.53 -7.46 9.88

0.038

7.90 7.31 4.07

NU

t-value

CE

Intercept Net PPE/TA log(TA) S&P Rating Borderline rating + rating - rating Investment grade border BBB- dummy BB+ dummy N -2 Log Likelihood Adj. R-square Panel B: Moody’s Intercept Net PPE/TA log(TA) S&P Rating Borderline rating + rating - rating Investment grade border BBB- dummy BB+ dummy N -2 Log Likelihood Adj. R-square

Est.

AC

Panel A: S&P

SC

RI

PT

Table 8 Use of Debt and Leases by Firms with “Borderline” Ratings. The dependent variable is the ratio of the value of operating leases to the sum of operating leases and total debt. In Panel A, the lease value is measured as the present value of the future minimum obligations (S&P method) discounted at the yield on a five year bond with the same rating as the firm. In Panel B, the value of operating lease is estimated using a multiple (Moody’s method). Loan maturity is measured in months. Borderline is a dummy variable equal to 1 if firms have S&P ratings with a “+” or “-” modifier, as reported by COMPUSTAT, 0 otherwise. “+ rating” is a dummy variable equal to 1 if firms have S&P ratings with a “+” modifier, 0 otherwise. “rating” is a dummy variable equal to 1 if firms have S&P ratings with a “-” modifier, zero otherwise. “Investment grade border” is a dummy variable equal to 1 if a firm has a credit rating of BBB- or BB+, 0 otherwise. “BBB-” is a dummy equal to 1 if a firm has a BBB- rating, zero otherwise. “BB+” is a dummy equal to 1 if a firm has a BB+ rating, zero otherwise. We assign the S&P ratings a numerical ranking in descending order starting at 26.

-0.44 4.35

12,397

12,523

0.044 0.028 12,523

0.016

0.019

0.019

0.012 0.01 0.002 0.001

23.13 2.74 -14.25 17.49

0.005 0.005

-1.75 3.41

0.286 0.031 -0.025 0.012

24.93 3.26 -14.47 17.50

0.041

7.54

0.286 0.032 -0.025 0.012

24.99 3.29 -14.44 17.36

7.39 3.39

12,397

12,397

12,523

0.052 0.027 12,523

0.0248

0.027

0.029

0.030

39

ACCEPTED MANUSCRIPT Highlights We examine the benefits of off-balance sheet operating leases in extending credit capacity.



We compare the effect of debt and leases on the cost of bank loans, public debt, and credit ratings.



Borrowing costs and credit ratings are less sensitive to lease financing than to debt financing.



This effect is stronger for financially constrained firms, or firms with lower marginal tax rates.



Firms close to ratings borderlines lease more, particularly around the investment grade border.

AC CE P

TE

D

MA

NU

SC

RI

PT



40