the sources of debt matter, too

The Sources Of Debt Matter, Too

Yang Liu*

University of Washington School of Business Administration

Department of Finance and Business Economics Box 353200

Seattle, WA 98195

[email protected]

This Draft: July 2004

* I thank Jarrad Harford, Jonathan Karpoff, Jennifer Koski, Paul Malatesta, and Edward Rice for helpful comments and suggestions.

2

The Sources of Debt Matter, Too

Abstract

This paper examines the effects of the use of outstanding bank loans, loans from non-

bank financial intermediaries, and unused bank lines of credit on a firm’s cash holdings,

equity risk, and investment. Firms with more outstanding bank loans have more cash

holdings and investment and a lower equity risk. Firms with more loans from non-bank

financial intermediaries have a lower equity risk and less investment. Moreover, firms with

more unused lines of credit have less cash holding, a lower equity risk, and more

investment. I also find that bigger and older firms and firms with more growth

opportunities have more unused lines than outstanding private debt. These results suggest

that not only the amount of debt but also the sources of debt matter in a firm’s financial

decisions.

3

I. Introduction

Bank loans and other privately placed debt are an important source of long-term funds

for U.S. corporations. The Federal Reserve System reports that $1,159 billion in bank

loans and $1,124 billion in other loans were outstanding in non-financial corporations at

the end of 20021. These amounts are roughly equal to these corporations’ public debt.

Bank loans and other privately placed debt are also special2. The existence of bank loans

and other private debt can serve as a signal of a borrower’s credit worthiness to the capital

market because financial intermediaries have more information than most other investors.

Banks and other financial intermediaries also monitor borrowers. Better-informed than

most other lenders, financial intermediaries have a comparative advantage in enforcing

debt contracts, and the concentrated ownership of private debt also mitigates the free-rider

problem in monitoring.

Several studies have examined the special functions of financial intermediaries from the

perspective of the valuation impact of private financing. For example, Mikkelson and

Partch (1985) and James (1987) find a positive stock abnormal return at the announcement

of bank credit agreements. Lummer and McConnell (1989) also find that the positive stock

abnormal return is associated with favorable bank loan renewals. However, current

empirical studies haven’t examined whether the uniqueness of private financing influences

a firm’s financial decisions.

The finance literature has also overlooked one important type of private debt – lines of

credit. Firms not only borrow term loans but also set up lines of credit with banks. For

example, 73% of my sample firms have unused lines of credit. Lines of credit also have

become more important recently than in the past. According to Bank of Japan, an

increasing number of companies are establishing committed lines of credit with financial

institutions to obtain funding3. However, as I know, none of the existing studies has

1 See Board of Governors of the Federal Reserve System (2002). 2 See Campbell and Kracaw (1980), Fama (1985), and Diamond (1984). 3 On May 4, 2004, Nikkei (Tokyo) reports.

4

documented the use of lines of credit in U.S firms or the differences between term loans

and lines of credit.

Private financing can affect a firm’s financial decisions beyond the effect of leverage

because of financial intermediaries’ monitoring function. First, borrowing from financial

intermediaries can decrease a firm’s optimal cash holdings and increase its investment

through the reduction of information asymmetry. Myers and Majluf (1984) suggest that

since information asymmetries increase a firm’s cost of external financing, firms with large

information asymmetries will hold more cash or have to underinvest when their internally

generated cash flows are not sufficient to finance all positive net-present-value projects.

Since financial intermediaries can evaluate a borrower’s project quality with more

information than most other investors, private borrowing can reduce a firm’s need to hold

cash and mitigate the underinvestment problem.

Second, borrowing from financial intermediaries can decrease a firm’s asset risk and

investment through the control of asset substitution. Jensen and Meckling (1976) argue that

levered firms have a higher asset risk than unlevered firms since shareholders in a levered

firm have an incentive to substitute risky assets for safe assets. Diamond (1989) shows that

monitoring from financial intermediaries and reputation effects constrain asset substitution

and reduce asset risk in levered firms. Diamond (1989) also implies that firms borrowing

more from financial intermediaries invest less in risky, negative net-present-value projects

than other firms when these firms have the same investment opportunities.

Different types of private financing can have distinctive effects. Fama (1985) and

Nakamura (1993) argue that the deposit relationships associated with commercial bank

borrowing facilitate the production of information. These deposit relationships, however,

are not present with non-bank loans. If information collected through deposit relationships

helps banks screen and monitor their borrowers, borrowing from banks will have stronger

effects on cash holdings, asset risk, and investment than borrowing from non-bank

financial intermediaries.

Unused bank lines of credit and outstanding private borrowing can also have different

effects. Financial intermediaries’ incentive to monitor may be weaker when there is unused

5

line of credit than when there are outstanding loans since no lending has occurred.

Financial intermediaries’ incentive to monitor may also be stronger when providing unused

lines of credit because borrowers may use lines of credit to invest in risky projects. The

differences in financial intermediaries’ incentive to monitor can lead unused lines of credit

and outstanding loans to affect a firm’s financial decisions differently.

I conduct several sets of tests to examine the effects of outstanding bank loans, non-bank

private debt, and unused bank lines of credit on a firm’s financial decisions. I use a unique

dataset that is composed of data hand collected from Moody’s industrial manuals and data

from COMPUSTAT and CRSP. The sample has 961 firms and 3,399 firm-year

observations over the 1996-2000 period. First, I examine the extent to which a firm’s

private borrowing changes its cash holdings. Second, I investigate the effect of private

borrowing on a firm’s equity risk. Third, I look directly at the relationship between the

amount of investment a firm makes and its private borrowing. With different regression

specifications, these tests show the differences between private debt and public debt, bank

debt and non-bank private debt, and outstanding private debt and unused bank lines of

credit.

The sources of debt have a significant effect on a firm’s financial decisions beyond the

effect of leverage. Firms with more outstanding bank loans have a lower equity risk and

more cash holdings and investment. Firms with more outstanding non-bank private debt

have a lower equity risk and less investment. In addition, firms with more unused bank

lines of credit have less cash holding, a lower equity risk, and more investment.

These results have the following implications. First, private borrowing differs

substantially from public borrowing. Private borrowing can mitigate both or at least one of

the information asymmetry and asset substitution problems. Second, firm-bank deposit

relationships facilitate the production of information. Both outstanding bank loans and

unused bank lines of credit increase investment through the reduction of information

asymmetry, while non-bank private debt does not. These results are consistent with the

argument that banks gain more information about firms through the firm-bank deposit

relationships. Third, banks have a strong incentive to monitor when firms have unused

6

bank lines of credit. Providing unused bank lines of credit to firms can be riskier for banks

than providing outstanding bank loans, since firms can use the funds for risky projects.

However, unused bank lines of credit still reduce a firm’s equity risk the same as

outstanding private does, which suggests that banks screen and monitor borrowers very

intensively when they provide lines of credit to firms.

When investigating the effects of private borrowing on a firm’s financial decisions, I

must deal with the endogeneity issue that arises from the selection of which firms borrow

from banks or non-bank financial intermediaries. Any observed effect may be a result of

reverse causality as firms receiving bank and non-bank private debt financing may have

little information asymmetry and moral hazard ex-ante. To understand the firm self-

selection process, I first examine the determinants of private borrowing. The use of

outstanding bank loans and non-bank private debt decreases with a firm’s size, age, and

growth opportunities, while the amount of unused lines of credit increases with growth

opportunities. In addition, I find that bigger and older firms and firms with more growth

opportunities have more unused lines of credit than outstanding private borrowing. The

determinants of private borrowing are used as instrumental variables to control for the

endogeneity problem.

This paper proceeds as follows. Section II develops testable hypotheses. Section III

introduces the sample and data. Section IV examines the determinants of private

borrowing. Section V reports the effects of private borrowing on a firm’s financial

decisions. Section VI concludes.

II. Hypotheses

Banks and other financial intermediaries have two functions that public investors do not

have. First, they have an information function. Several theoretical papers, e.g. Campbell

and Kracaw (1980), Fama (1985), and Boyd and Prescott (1986), argue that financial

intermediaries have more information than other capital market participants and their

lending decisions reveal a borrower’s credit worthiness. Second, financial intermediaries

7

have a monitoring function. Diamond (1984) shows that the costly monitoring function is

delegated to financial intermediaries since the concentrated ownership of private debt

avoids the free-rider problem in monitoring and enforcing debt contracts and minimizes the

cost to produce information useful to resolve the problem.

Aoki (1994) classifies these two functions into three stages. The primary distinction

among the three stages of monitoring is the timing of an action taken by the supplier of

funds in relation to the transfer of funds from the investor to the firm. The first stage of

monitoring, ex ante, refers to the investor’s assessment of the credit-worthiness of

investment projects proposed by corporate firms and their screening. The second stage,

interim monitoring, refers to an investor checking the ongoing behavior of management

and the operation of the firm in general, and the use of funds in particular, after the funds

are committed. The third kind of monitoring, ex post, refers to the verification of

performance outcome of the firm, judgment on the long-run viability of the firm in case of

financial distress, and the use of that information for possible corrective or punitive

actions.

Based on different stages of financial intermediaries’ monitoring function, I develop two

hypotheses, the information asymmetry hypothesis and the moral hazard hypothesis. The

information asymmetry hypothesis focuses on the effects of ex ante monitoring on

financial decisions. The moral hazard hypothesis focuses on the effects of interim and ex

post monitoring. The three stages of monitoring and the focuses of the information

asymmetry hypothesis and moral hazard hypothesis are shown in Figure 1.

0 1 2 3

ex ante invest interim monitoring ex post

information asymmetry hypothesis moral hazard hypothesis

Figure 1 Monitoring functions of financial intermediaries

8

2.1 Information asymmetry hypothesis

The information asymmetry hypothesis, based on Myers and Majluf (1984), predicts that

firms with more private debt hold less cash and invest more than other firms. Based on

Fama (1985), the information asymmetry hypothesis also predicts that firms borrowing

from banks hold less cash and invest more than firms borrowing from non-bank financial

intermediaries.

Information asymmetry between insiders and outside investors increases the cost of

external financing and causes underinvestment. Leland and Pyle (1977) argue that since

only managers know the quality of the project, the yield on borrowings will reflect only the

average project quality. If the market were to place an average value greater than average

cost on projects, the potential supply of low quality projects may be very large, since

managers could foist these upon the uninformed market and make a sure profit. But this

argues that the low average project quality leads to the consequence that even projects that

are known by managers to merit financing cannot be undertaken due to the high cost of

capital. Myers and Majluf (1984) consider a firm that must issue external securities to raise

cash to undertake a valuable investment project. When uninformed investors in the capital

market undervalue the firm’s security, the external financing cost is so high that the firm

will underinvest. Their paper suggests that because of information asymmetry, firms have

the tendency to rely on internal sources of funds and to underinvest when internally

generated cash flows are insufficient to finance all positive net-present-value projects.

Financial intermediaries have an advantage over public investors in collecting

information. Campbell and Kracaw (1980) argue that financial intermediaries emerge as

information producers because the production of information, the protection of

confidentiality, the provision of transactions services, as well as other intermediary

services, are naturally complimentary activities. Fama (1985) draws a distinction between

outside debt and inside debt. He claims that bank loans are inside debt as banks can gain

access to information from a borrower’s decision-making process that would not otherwise

9

be publicly available. Ramakrishnan and Thakor (1984), Boyd and Prescott (1986), and

Berlin and Loeys (1988) also contend that private lenders have an information advantage

over lenders in the public debt markets. While public lenders have only public information

to assess the risks of a firm, financial intermediaries have the access to non-public

information, sometimes regarding the future potential of the firm.

Empirical findings also support banks’ information advantages. Mikkelson and Partch

(1985) and James (1987) find a significant positive stock abnormal return at the

announcement of credit agreements with banks. Krishnaswami, Spindt and Subramaniam

(1999) also show that firms with larger potential information asymmetries use more

privately placed debt than other firms.

Since financial intermediaries have an information advantage over other investors, firms

borrowing more from financial intermediaries can hold less cash and invest more than

other firms. Since financial intermediaries have more information about borrowers and

their projects, borrowing from financial intermediaries can mitigate the adverse selection

problem. For example, for some positive net-present-value projects, financial

intermediaries are willing to lend the firm at a lower interest rate that reflects the quality of

the project accurately than the interest rate charged by public investors, at which these

projects would be foregone4. As a result, borrowing from financial intermediaries

alleviates the underinvestment problem. The reduction of adverse selection through private

borrowing also releases the pressure for firms from holding a lot of cash. Since private

debt can be borrowed at an interest rate closely reflecting a project’s quality, firms can rely

less on internally generated cash flows to finance their positive NPV projects. Instead, they

can distribute cash to shareholders through dividends and stock repurchase.

Alternative to the negative effect of private financing on cash holdings predicted by the

information asymmetry hypothesis, private borrowing can also have a positive effect on a

firm’s cash holdings. Smith and Warner (1979) argue that private debt usually has more

restrictive covenants in restricting dividend payout, various financial ratios, and additional

4 When there is a big information asymmetry between managers and outside investors, there won’t be any lending from uninformed investors to the firm. The interest rate that uninformed investors charge reflects the firm’s average project quality, and at that rate managers will invest in only bad projects. Thus, at equilibrium, there is no lending from uninformed investors.

10

debt issues than public debt does. Such restrictions limit a borrower’s financial flexibility.

In a survey of 392 CFOs about the cost of capital, capital budgeting, and capital structure,

Graham and Harvey (2001) report that financial flexibility is a firm’s top concern when

issuing debt. Therefore a competing argument for the information asymmetry hypothesis is

that firms, in order to minimize the negative consequences of debt covenants on their

investments and other operations, will hold more cash as they borrow more private debt.

The difference between banks and non-bank financial intermediaries can affect the

production of information. Fama (1985) and Nakamura (1993) argue that the deposit

relationships associated with commercial bank borrowing facilitate the information

production and monitoring, and these relationships are not present with non-bank private

borrowing. However, the evidence on whether banks and non-bank financial intermediaries

have the same functions are mixed. Studying private placement markets, Carey, Prowse,

Rhea, and Udell (1993) show that firms borrowing from non-bank financial intermediaries

have less information asymmetry than firms borrowing from banks. James (1987) finds a

significant negative stock price response at the announcement of private placements used

to repay bank loans. However, James and Wier (1990), in contrast, find no difference

between bank debt and non-bank debt in certifying firms in their initial public equity

offerings.

Bank borrowings can have a stronger effect in reducing cash holdings and increasing

investment than non-bank private borrowings. If deposit relationships provide more private

information, such as daily cash flows, than information provided by a firm when it seeks

financing from financial intermediaries, banks will have a comparative advantage in

screening potential borrowers over non-bank financial intermediaries. Therefore,

borrowing from banks can reduce a firm’s cash holdings and increase its investment more

than borrowing from non-bank financial intermediaries.

2.2 Moral hazard hypothesis

11

The moral hazard hypothesis, which is based on Diamond (1989), predicts that firms

borrowing more from financial intermediaries have a lower asset risk and less investment

than other levered firms. Moreover, based on the assumption that deposit relationships

between banks and firms provide more information to facilitate monitoring, the moral

hazard hypothesis also predicts that firms borrowing from banks have a lower asset risk

and less investment than firms borrowing from non-bank financing intermediaries.

One moral hazard problem between debt holders and shareholders is asset substitution,

which arises from the adverse incentives of limited liability. Jensen and Meckling (1976)

argue that shareholders in a levered firm have an incentive to undertake riskier projects

because they have unbounded upside potential for future cash flows but face only bounded

downside potential due to limited liability. Viewing a levered firm’s equity as a call option

on the firm’s underlying assets, Galai and Masulis (1976) argue that substituting assets of

high risk for those of low risk can increase the volatility of the firm’s assets, thus increase

the share value. However, asset substitution decreases the value of debt.

The moral hazard problem can be partially prevented by writing restrictive debt

covenants, which increases the costs of asset substitution to shareholders. For instance,

Smith and Warner (1979) argue that restrictions on sale, lease or disposal of any

substantial part of its properties and assets; and on the collateral of assets, limit a firm’s

ability to engage in asset substitution. Private lenders, having a comparative advantage in

writing and enforcing bond covenants, can effectively reduce the agency costs due to moral

hazard. However, since a debt contract cannot specify all contingencies, covenants can,

only to some extent, solve the moral hazard problem.

Private borrowings can also constrain asset substitution in a levered firm through

monitoring. Diamond (1989) shows that monitoring from financial intermediaries and the

cost of losing reputation jointly eliminate the conflict of interests between borrowers and

lenders about the choice of risk in investment decisions. His model suggests that firms

borrowing from financial intermediaries engage in less asset substitution and have a lower

asset risk than firms borrowing from the public debt market.

12

Through the control of asset substitution, borrowing from financial intermediaries also

affects the amount of investment a firm will make. Assume that there are two firms, firm A

and B. They have the same investment-opportunity sets, which consist of both safe,

positive net-present-value projects and risky, negative net-present-value projects. Firm A

borrows from financial intermediaries, while firm B borrows from the public debt market.

Financial intermediaries perform interim monitoring after they lend money to firm A and

ex post monitoring after the outcome of the projects invested in is realized. With a credible

punishment of losing reputation through financial intermediaries’ monitoring, Diamond

(1989) implies that firm A will invest only in positive net-present-value projects. Firm B,

on the other hand, will invest in not only all the positive net-present-value projects but also

some negative net-present-value projects. In terms of the amount of investment, Diamond

(1989) suggests that, controlling for investment opportunities, firms with more private debt

invest less than other levered firms5.

Because of their deposit relationships with firms, banks can be more efficient in

monitoring. Financial intermediaries collect information from their borrowers to enforce

covenants and monitor cash flows. If deposit relationships provide banks information

beyond what is revealed by their borrowers, banks are more likely to detect asset

substitution than non-bank financial intermediaries. As a result, firms borrowing from

banks will have a lower asset risk and less investment than firms borrowing from non-bank

financing intermediaries.

2.3 Summary of hypotheses

Focusing on different stages of financial intermediaries’ monitoring function, the

information asymmetry hypothesis and the moral hazard hypothesis have the following

predictions.

5 The prediction that firms with more private debt have less investment is based on the assumption that monitoring from financial intermediaries does not change firms’ incentive to forego positive NPV projects (the debt overhang problem in Myers (1977)). Because of information asymmetry between managers and outside investors on potential projects that a firm can undertake, it’s almost impossible for financial intermediaries to force the firm to invest in all positive NPV projects.

13

The effect of private debt on Information asymmetry

hypothesis

Moral hazard

hypothesis

Cash holdings Negative N/A

Asset risk N/A Negative

Investment Positive Negative

Based on the assumption that the deposit relationships between firms and banks

facilitate the production of information, bank borrowing will have stronger effects than

non-bank private borrowing on cash holdings, asset risk, and investment.

III. Sample and Data Description

3.1 Sample

My dataset merges the COMPUSTAT annual industrial file, the COMPUSTAT research

annual industrial file, the Center for Research in Security Prices (CRSP) tapes, and

Moody’s industrial manuals. The data on the amount of outstanding bank loans, loans from

non-bank financial intermediaries, and unused bank lines of credit are hand collected from

Moody’s industrial manual. Other financial data are collected from COMPUSTAT, and the

stock return data are collected from CRSP. I restrict sample firms to manufacturing firms

(SIC codes between 2000 and 3999) to avoid concerns with regulation. All sample firms

are traded on NYSE or AMEX because Moody’s industrial manuals have coverage only on

such firms. I first select all manufacturing firms that are covered by COMPUSTAT in any

year from 1996 to 2000. There are 996 such firms or 4,200 firm-year observations. Then

only 961 firms or 3,399 firm-year observations have private debt data in Moody’s

industrial manuals from 1995 to 1999. Table 1 presents the summary statistics for variables

used in this study.

3.2 Dependent variables

14

There are three dependent variables: cash holdings, equity risk, and investment. I

measure cash holdings by the ratio of cash and marketable securities to the value of total

assets. I use the standard deviation of daily common stock returns over each year to

measure equity risk. Equity risk does not only incorporate asset risk but also a firm’s

financial risk. Investment is measured by net investment at t=1 divided by the book value

of fixed assets in year 0. Net investment is defined as capital expenditure minus

depreciation.

3.2 Private borrowings

There are two measures of outstanding private debt: outstanding bank loans and non-

bank private debt. Both are scaled by total assets6. Most firms have a combination of bank

loans, loans from other financial intermediaries, and public debt. Outstanding bank loan is

measured as the ratio of outstanding long-term bank debt to total assets. Bank loans

include revolving and non-revolving loans, and term loans outstanding. 68% of sample

firms (651 firms) have outstanding bank loans. On average, 22.2% of sample firms’ debt is

borrowed from banks. The non-bank private debt is the ratio of loans placed with non-bank

financial intermediaries to total long-term debt. Non-bank financial intermediaries include

insurance companies, pension funds, mutual funds, and other financing firms. 25% of

sample firms (238 firms) have outstanding loans from non-bank financial intermediaries,

and the mean of the non-bank private debt ratio is 7%. Among 961 sample firms, 712 firms

(74%) have outstanding loans from banks or non-bank financial intermediaries or both, and

the amount of outstanding private debt accounts for 29.3% of sample firms’ total debt.

6 In all prior studies on private debt, the amount of private debt is scaled by total long-term debt to represent the mix of public and private debt. However, since firms holding unused bank lines may not have outstanding long-term debt, using total long-term debt to scale will result in loss of observations. In order to compare the determinants of outstanding private debt and unused bank lines of credit, these dependent variables are scaled by total assets. Scaling outstanding private debt by the amount of long-term debt yields the same regression results.

15

Another measurement of private borrowing is unused bank lines of credit. Firms often

set up a line of credit with banks and do not borrow up to the full amount allowed under

their bank credit agreements. Liu (2004) reports that firms keep unused lines of credit for

temporary working capital needs, dividend payment, acquisition, or emergency external

capital needs. A bank credit agreement usually specifies the maximum amount of money

that a firm can borrow from the bank. The amount of the unused lines of credit is the

difference between the maximum amount that a firm can borrow under credit agreements

and the amount of outstanding loans.

Although unused lines of credit are not debt, they can affect a firm’s financial decisions

as well. No matter whether banks provide term loans or lines of credit to firms, banks need

to check the credit worthiness of the borrowers. Therefore, unused lines of credit will also

reduce cash holdings and increase investment. The motivation to monitor borrowers can be

different when banks commit to lines of credit and when they provide term loans. Since

unused lines of credit are not debt yet, banks may have a weaker incentive to monitor.

Banks’ incentive to monitor can also be stronger when committing lines of credit. Liu

(2004) reports that firms hold lines of credit for emergency external capital needs. As a

result, unused lines of credit give firms the chance to invest in risky projects. To prevent

asset substitution, banks will monitor intensively.

Firms hold a significant amount of unused lines of credit. The average amount of unused

bank lines of credit held by the sample firms represents 7.1% of these firms’ total assets.

406 firms have data for each year over the 5-year sample period. Table 2 reports that

among these firms, 62 firms (15.3% of 406 sample firms) do not have an unused line of

credit; 64 firms (15.8%) have an unused line of credit over one year; 81 firms (20.0%)

have an unused line of credit over two years; 84 firms (20.7%) have an unused line over

three years; 73 firms (18.0%) have an unused line of credit over 4 years; and 42 firms

(10.3%) have an unused line of credit over all 5 years in the sample period.

Most firms have borrowing relationships with financial intermediaries. Table 3 reports

that 848 firms (88% of the sample firms) have borrowing relationships with either banks

or non-bank financial intermediaries. 88 firms have only outstanding bank loans, and these

16

firms hold 14.4% bank debt, as a percentage of total assets. Only 28 firms (3%) have

borrowing from non-bank financial intermediaries but not from banks. 136 firms (14.2%)

have unused bank lines of credit but no outstanding private debt. A significant portion of

sample firms (40.2%) has both outstanding bank loans and unused bank lines of credit but

no outstanding non-bank private debt. 142 firms have all three types of private debt.

3.3 Control variables

Following Barclay and Smith (1995), I use the market-to-book ratio, which is the ratio

of the market value of total assets to the book value of total assets, as a proxy for the

present value of a firm’s growth opportunities. Growth options increase a firm’s market

value relative to its book value since intangible assets such as growth options are not

included in the book value of assets. I estimate the market value of assets as the book value

of assets minus the book value of equity plus the market value of equity. The market-to-

book ratio is included in the regressions of cash holdings and investment. Since firms with

a high market-to-book ratio are more likely to be cash constrained, they will hold more

cash. Investment should also increase with the market-to-book ratio. The market-to-book

ratio has a mean of 1.842.

The natural log of sales, which are stated in 1996 dollars, measures firm size. Firm size

is controlled for in the regressions of cash holdings and equity risk. Cash holding/total

assets is expected to decrease with a firm’s size. Because of the fixed cost of outside

funding, big firms with a larger issue size will, on average, find it more cost-effective to

issue new securities. Having a lower cost of cash shortage, they will have a lower ratio of

cash holding to total assets than small firms. Big firms with more diversified operations

will also have a lower asset risk than small firms.

A firm’s debt level also affects its cash holdings, equity risk, and investment. Opler,

Pinkowitz, Stulz, and Williamson (1999) argue that leverage is a proxy for the degree to

which the capital markets monitor a firm and that managers in firms with low debt are

likely to hold more cash for their own objectives. The leverage ratio also affects a firm’s

17

equity risk by introducing financial risk to the firm. Moreover, Lang, Ofek, and Stulz

(1996) find that leverage negatively affects the investments made by firms whose growth

opportunities are either not recognized by the capital markets or are not sufficiently

valuable to overcome the effects of their debt overhang. Therefore, leverage, measured by

the ratio of the book value of short-term and long-term debt to the book value of total

assets, is included in the regressions of cash holdings, equity risk, and investment.

Leverage has a mean of 11.7%.

Cash flow is measured by earnings before interest and depreciation, but after taxes,

scaled by total assets. Fazzari, Hubbard, and Petersen (1988), Hoshi, Kashyap, and

Scharfstein (1991), Lang, Ofek, and Stulz (1996) and others show that investment is

related to the availability of internal funds. Since cash flow net of interest expense partially

captures the effects of leverage, I use a cash flow measure gross of interest as in Lang,

Ofek, and Stulz (1996).

Firms with higher research and development expenses (R&D) will hold more liquid

assets. When information asymmetries are important, a cash flow shortfall forces firms to

contract investment. Since investment in research and development is subject to a larger

degree of information asymmetry than other types of investment, it’s more difficult to get

external financing. As a consequence, firms with more R&D expense will hold more cash.

Net working capital, measured by working capital minus cash scaled by total assets, is a

substitute for cash. Firms may choose to insure themselves against losses by holding liquid

assets besides cash.

Lagged capital expenditure scaled by fixed assets can be related to a firm’s current

investment. For example, some investment projects take several years to complete, so there

can be a positive correlation between lagged capital expenditure and current investment.

On the contrary, it’s unlikely for firms that just invested in machinery last period to invest

it again this period, so lagged capital expenditure can be negatively correlated with current

investment.

IV. Determinants of Private Borrowing

18

In investigating the effects of private borrowing on a firm’s financial decisions, it is

essential to address the endogeneity of the selection of who borrows from financial

intermediaries in the first place. Even if bank debt or non-bank private debt has no effect

on cash holdings, asset risk or investment, we may observe a significant relation between

these variables if firms that receive bank and non-bank private debt financing are those that

are less likely to have the information asymmetry and asset substitution problem. For

example, suppose that monitoring does not affect a firm’s asset risk and that financial

intermediaries lend money only to firms that are less likely to substitute risky assets for

safe assets. The selection of which firms borrow from financial intermediaries will result in

a negative correlation between asset risk and private borrowing and lead to the conclusion

that private borrowing controls asset substitution through monitoring. Therefore, unless

controlling for the endogeneity of private debt financing, we cannot make any assertion on

the causality between the use of private debt and financial decisions for any significant

correlation observed between them. In this section, I first examine the determinants of the

use of outstanding bank loans, non-bank private debt, and unused bank lines of credit.

These determinants are used as instrumental variables in later regressions to control for the

endogeneity problem.

Previous studies suggest that the use of private debt should be related to flotation costs

of new debt issues, information asymmetry, and moral hazard. Issuing public debt costs

more than issuing private debt since public issues are associated with investment banker

fees, filing and legal fees, and other transaction costs7. As a result, small firms whose debt

issue size is generally small will use more private debt than big firms. Firms with large

information asymmetry and severe moral hazard problems will also use more private debt

than other firms. As discussed in section II, financial intermediaries have a comparative

advantage in collecting information and monitoring over other investors, therefore

borrowing from financial intermediaries can reduce the debt contract costs arising from

information asymmetry and moral hazard.

7 See Smith (1986) and Blackwell and Kidwell (1988).

19

The following regression is used to analyze the cross-sectional differences in the use of

private debt.

PBit = α + Xit*Γ + εit (1)

In equation (1), PB is private borrowing – outstanding bank loans, non-bank private debt,

or unused lines of credit. Xit are dependent variables, which include a firm’s size, age,

market-to-book ratio, fixed-asset ratio, and residual standard deviation. Tobit regressions

are used to estimate the coefficients since private debt data are censored8. Each variable is

measured at its time-series mean within each firm.

Size is used as a proxy for flotation costs. Since the size of debt issues is positively

correlated with a firm’s size, the use of private debt will be negatively related to a firm’s

size. Size is also a proxy for information asymmetry, since bigger firms are better known

by investors than small firms.

Age and the residual standard deviation are used as proxies for information asymmetry.

The age of a firm is defined as the number of years since its incorporation. The data on the

year of a firm’s incorporation is obtained from Moody’s Industrial Manual. The residual

standard deviation is defined as the standard deviation of the residuals from the regressions

of daily stock returns in the previous year on the market returns. Bhagat et al. (1985),

Blackwell et al. (1990), Dierkens (1991), and Krishnaswami et al. (1999), argue that there

is large information asymmetry when managers have a relatively large amount of value-

relevant, firm-specific information that is not shared by the market. Until this information

is revealed to the market, investors bear some firm-specific uncertainty. If investors in the

market and the manager of a firm are equally well informed about the market-wide

(systematic) factors influencing firm value, residual volatility in a firm’s stock returns may

be used as a proxy for information asymmetry about firm-specific information.

The market-to-book ratio and fixed-asset ratio are used as proxies for the moral hazard

problems. Barclay and Smith (1995) argue that firms with more growth opportunities have

8 Private borrowing data are censored because we observe no private borrowing when (α + Xit*Γ + εit) is negative.

20

more severe incentive problems since shareholders of high growth firms can substitute

riskier projects for safe ones more easily. Therefore, firms with a higher market-to-book

ratio and a lower fixed-asset ratio are expected to use more private debt to reduce debt

contract costs.

4.1 The use of different types of private debt

The use of outstanding bank loans decreases with a firm’s size, age, market-to-book

ratio, and fixed-asset ratio, and increases with the residual standard deviation. In the first

regression reported in Table 4, the coefficient on size is –0.130, and the coefficient on age

is –0.202. Both coefficients are statistically significant at the 1% level. The residual

standard deviation has a positive and statistically significant coefficient. These results are

consistent with the flotation argument and the information asymmetry argument. The

coefficients on the market-to-book ratio and the fixed-asset ratio are negative, but not

statistically significant. The negative coefficient on the market-to-book ratio suggests that

there is a hold-up problem associated with bank borrowing. Rajan (1992) argues that

information acquired by banks as a part of an ongoing relationship can create an

information monopoly or a hold-up problem that it’s costly for a borrower to switch

lenders. He also argues that such problems are likely to be particularly acute for firms with

valuable investment opportunities.

The use of non-bank private debt decreases with a firm’s size, age, market-to-book ratio,

and fixed-asset ratio, and increases with the residual standard deviation. However, only the

coefficients on size and age are statistically different from zero at the 10% level.

The holding of unused lines of credit decreases with a firm’s size, age, fixed-asset ratio,

and residual standard deviation, and increases with the market-to-book ratio. Coefficients

on all variables except the fixed-asset ratio are statistically different from zero at the

conventional level. Different from the use of outstanding private debt, the amount of

unused lines of credit increases with the market-to-book ratio. The amount of unused lines

of credit also decreases with the residual standard deviation, which suggests that banks are

21

more likely to commit lines of credit to firms with a lower risk to reduce the chance of

asset substitution.

4.2 Differences in the use of different types of private borrowing

Regressions in Table 5 examine the differences among the use of different types of

private borrowing. Regressions in Table 4 show that there are differences in the

determinants of outstanding bank loans, non-bank private debt, and unused lines of credit,

however we cannot make formal statistical inferences on the differences based on those

results. Regressions in Table 5, using outstanding bank loans/outstanding private debt and

unused lines of credit/the sum of outstanding and unused borrowing from financial

intermediaries as dependent variables, directly analyze the differences between the use of

outstanding bank loans and non-bank private debt and between the holding of unused lines

and outstanding private borrowing.

There are some differences between the use of outstanding bank loans and non-bank

private debt, but the differences are not statistically significant. The first regression

reported in Table 5 shows that bigger and younger firms and firms with a higher market-to-

book ratio, fixed-asset ratio, and residual standard deviation use more outstanding bank

loans than non-bank private debt. However, none of the coefficients in the regression is

statistically different from zero.

Bigger and older firms and firms with a higher market-to-book ratio, and a lower fixed-

asset ratio and residual standard deviation have more unused lines of credit relative to

outstanding private debt. In the second regression in Table 5, the coefficient on size is

0.043, statistically significant at the 1% level; and the coefficient on age is 0.040,

statistically significant at the 10% level. The coefficient on the market-to-book ratio is

0.085, statistically significant at the 1% level. Liu (2004) report that firms may hold

unused lines of credit to keep down their borrowings while ensuring that they can quickly

secure funds in an emergency, such as a sudden deterioration in the issuance environment

for corporate bonds. Since it’s more costly for firms with good investment opportunities to

22

underinvest, firms with a high market-to-book ratio hold more unused lines of credit than

firms with a low market-to-book ratio. The market-to-book ratio may also represent the

quality of the management in a firm. Then the positive relation between the unused lines of

credit and the market-to-book ratio also implies that banks are more willing to commit to

lines of credit to firms of a better quality. Banks are unable to evaluate borrowers’ project

quality before they commit lines of credit; therefore banks are more likely to provide lines

of credit to firms with a good management team in order to ensure the project quality. The

coefficients on the fixed-asset ratio and the residual standard deviation are negative but not


My findings suggest that firms use different types of debt in different stages of their life

cycle. When firms are small and young, they borrow mainly from banks and non-bank

financial intermediaries. As firms grow bigger and older, they still borrow from banks,

however they are more likely to hold lines of credit for their investment opportunities in

case of an emergency than have outstanding bank loans. After firms have established

reputation in the financial market, they switch from financial intermediaries to public

investors when they want to raise external capital.

The determinants of the use of outstanding bank loans, non-bank private debt, and

unused lines of credit will be used as instrumental variables in later tests to control for the

endogeneity issue associated with private financing, as long as they do not belong to the

regressions of cash holdings, equity risk, or investment themselves.

V. Effects of Private Borrowing

5.1. Cash holdings and private borrowing

5.1.1 Regression models

To investigate the effects of private borrowing on cash holdings, I extend the regression

used in Opler et al. (1999) to include different types of private debt as testing variables.

For each test, I have three regression specifications.

23

CHit = α + β1*BLit + β1*NBPDit + β1*ULCit + Xit*Γ+ εit (2)

CHit = α + β1*TBDit + β2*NBPDit + Xit*Γ+ εit (3)

CHit = α + β1*OPDit + β2*ULCit + Xit*Γ+ εit (4)

In equation (2), (3), and (4), CH is cash holding/total assets; BL is outstanding bank

loans; NBPD is the non-bank private debt; ULC is unused lines of credit; TBD is total

bank debt; OPD is total outstanding private debt; and X is the vector of control variables

that include a firm’s market-to-book ratio, size, leverage ratio, cash flow, research and

development expenses, and net working capital.

In equation (2), outstanding bank loans, non-bank private debt, and unused bank lines of

credit are included in the regression separately. This specification assumes that three types

of private borrowing have different effects.

In equation (3), non-bank private debt and all bank debt, which is the sum of outstanding

bank loans and unused bank lines of credit, are included as testing variables. This

specification assumes that the sources of debt matter in the cash holding decisions while

whether lending has taken place does not.

In equation (4), unused lines of credit and outstanding private debt, which is the sum of

outstanding bank loans and non-bank private debt, are included as dependent variables.

This last specification emphasizes whether lending has been committed but not the sources

of funds.

Each variable is measured at its time-series mean within each firm. There are two

reasons for measuring these variables at the time-series mean. First, the predictions from

the information asymmetry hypothesis and the moral hazard hypothesis are about

differences across firms. For example, the information asymmetry hypothesis predicts that

firms with more unused bank lines of credit invest more than other firms since bank

borrowing reduces information asymmetry. In exploiting the time-series differences, we

24

may observe that firms draw from the line of credit to make investment. The decrease in

the unused bank lines of credit and the increase in investment lead to a negative correlation

between them. Therefore, the time-series analysis is not appropriate to test the information

asymmetry and moral hazard hypotheses. The cross-sectional regressions preserve the

dispersion across firms, but exploit no time-series variation in the observations.

Second, using cross-sectional data avoids the weak instrumental variable problem.

Instrumental variables that are not correlated with the error terms in the regressions are

difficult to identify since most financial decisions are interrelated. When excluded

exogenous variables do not have a strong explanatory power for measures of private

borrowings (the weak instrumental variable problem), the bias in the OLS estimators is

exacerbated in the two-stage estimation and the coefficient estimates are not consistent.

The determinants of private borrowing identified in the previous section have a stronger

power in explaining the cross-sectional differences in the use of private debt than in

explaining the cross-sectional and time-series differences9. Therefore, using cross-sectional

regressions also has a better fit in the two-stage estimations with instrumental variables.

I use the generalized method of moments (GMM) to estimate the coefficients. There are

other estimation methods compatible with the instrumental-variable approach, such as the

two-stage least square estimation and the limited information maximum likelihood

estimator. However, only GMM can correct for the heteroskedasticity in the error terms –a

common issue when using cross-sectional data. Moreover, using GMM enables me to test

the validation of the model specification when the number of excluded exogenous

variables is more than the number of endogenous variables. The over-identification test

tests whether the regression model is correctly specified by examining the correlation

between the excluded exogenous variables and the regression error terms.

5.1.2 Correlation analysis

9 The excluded exogenous variables are jointly different from zero at 1%, 5%, or 10% level in cross sectional regressions of outstanding bank loans, non-bank private debt, and unused bank lines of credit. They are less statistically significant in the time-series cross-sectional regressions of non-bank private borrowing.

25

Table 6 presents the correlation analysis of the variables used in the regressions of cash

holdings. Consistent with the information asymmetry hypothesis, cash holdings have a

negative and statistically significant coefficient with all three private debt variables –

outstanding bank debt, non-bank private debt, and unused lines of credit. Cash holdings

also have a positive correlation with a firm’s market-to-book ratio and research and

development expense, and a negative correlation with a firm’s size, leverage ratio, cash

flow, and net working capital. All these correlations are statistically significant at the 1%

level.

5.1.3 Regression results

The first three regressions in Table 7 show that using outstanding bank loans, non-bank

private debt, and unused bank lines of credit all reduces a firm’s cash holdings. The

coefficients on all testing variables are negative and statistically significant at the

conventional level. The results are consistent with the information asymmetry hypothesis

that borrowing from financial intermediaries mitigates the information asymmetry

problem.

Coefficients on most control variables are consistent with the results in Opler,

Pinkowitz, Stulz, and Williamson (1999). A firm’s cash holdings increase with its market-

to-book ratio and research and development expenses, and decrease with its size, leverage

ratio, and net working capital. Including private debt variables in the regressions increases

the explanatory power of the regressions. The adjusted R2 of the cross-sectional regression

reported by Opler et al (1999) is 0.381, while the adjusted R2 of my regressions is about

0.47.

Although the findings are consistent with the information asymmetry hypothesis, the

coefficient estimates in the first three regressions can be biased due to the endogeneity

problem discussed in Section IV. Regression 4, 5, and 6 in Table 7, repeat the same

analysis and use instrumental variables to control for the endogeneity problem. Excluded

exogenous variables include a firm’s age, fixed-asset ratio, size2, and residual standard

26

deviation. Instrumental variables are also measured at the time-series mean within each

firm.

Regression 4, 5, and 6 show that using outstanding bank loans increases a firm’s cash

holdings while unused lines of credit decreases a firm’s cash holdings. In regression 4, the

coefficients on bank loans and unused lines of credit are 0.46 and –0.49, respectively. Both

coefficients are statistically significant at the 1% level. The unused line of credit also has a

negative and statistically significant coefficient of -0.618 in the last regression. The

coefficients on non-bank private debt are negative in regression 4 and 5, and not


Regression results suggest that holding unused lines of credit is equivalent to holding

cash since the lines are available for firms’ immediate use. However, one dollar of lines of

credit is worth less than a dollar of cash, since there is cost associated with holding lines of

credit. Outstanding bank loans, on the other hand, make firms to hold more cash since the

debt covenants constrain a firm’s financial flexibility.

5.2 Equity risk and private borrowing

5.2.1 Regression model

The following regressions of equity risk test the moral hazard hypothesis that borrowing

from financial intermediaries constrains firms from asset substitution and reduces firms’

asset risk.

ERit = α + β1*BLit + β1*NBPDit + β1*ULCit + Xit*Γ+ εit (5)

ERit = α + β1*TBDit + β2*NBPDit + Xit*Γ+ εit (6)

ERit = α + β1*OPDit + β2*ULCit + Xit*Γ+ εit (7)

27

Equity risk, the dependent variable used in the above regressions, does not only

incorporate a firm’s asset risk of but also its financial risk. Xit is the vector of control

variables that include a firm’s size and leverage ratio which controls for the financial risk

of the firm. All variables are measured at their time-series means within each firm. GMM

estimators are used.


Table 8 reports the correlation analysis of the determinants of equity risk. Equity risk is

positively correlated with outstanding private debt. The correlation between equity risk and

bank debt is 0.101, statistically significant at the 1% level; and the correlation between

equity risk and non-bank private debt is 0.082, statistically significant at the 5% level.

These positive correlations are contradictory to the moral hazard hypothesis. Moreover,

equity risk and unused lines of credit have a correlation of –0.134, statistically significant

at the 1% level. In addition, equity risk is negatively correlated with a firm’s size and

positively correlated with a firm’s leverage ratio.


Outstanding borrowing reduces a firm’s equity risk. Without controlling for the

endogeneity problem, coefficients on all outstanding private debt variables in the first three

regressions are negative. However, none of them is statistically different from zero. When

the fixed asset ratio, residual standard deviation, market-to-book ratio, and size2 are used as

instrumental variables10 in the last three regressions, coefficients on bank debt, non-bank

private debt, and all outstanding private debt are still negative and become statistically

significant. The regressions also show that bigger firms and less levered firms have a lower

equity risk.

Having unused bank lines of credit also decreases a firm’s equity risk as outstanding

bank loans do. In Table 9, the coefficients on unused lines of credit in the first and third

10 Regression results are not sensitive to the choice of instrumental variables. For example, dropping the residual standard deviation from the instrumental variable does not change the results in regression 4, 5, and 6.

28

regressions are –0.017, and statistically significant at the 1% level. The coefficient on all

bank debt in the second regression is –0.010, also statistically significant at the 1% level.

In the regressions controlling for the endogeneity problem, the coefficients on the unused

lines of credit remain negative and statistically significant.

The result implies that banks monitor borrowers intensively when there are unused bank

lines of credit. Unused bank lines of credit can be subject to a more severe asset

substitution problem than outstanding private debt is, since banks are unable to evaluate

the project quality before committing lines of credit. However, holding used lines of credit

still reduces a firm’s equity risk as outstanding private debt does. Therefore, banks are very

careful in screening and monitoring firms when providing lines of credit.

5.3 Investment and private borrowing

5.3.1 Regression model

Since the information asymmetry hypothesis and the moral hazard hypothesis have

opposite predictions on the impact of private borrowing on a firm’s investment, the

following regressions test which of the information and monitoring function of banks and

non-bank financial intermediaries is more prominent.

Iit = α + β1*BLit + β1*NBPDit + β1*ULCit + Xit*Γ+ εit (8)

Iit = α + β1*TBDit + β2*NBPDit + Xit*Γ+ εit (9)

Iit = α + β1*OPDit + β2*ULCit + Xit*Γ+ εit (10)

In equation (8), (9), and (10), Iit is investment; Xit is the vector of control variables that

include cash flow, the market-to-book ratio, lagged capital expenditure, and the leverage

ratio. All variables are measured at their time-series means within each firm. GMM

estimators are used.

29


Table 10 provides the correlation analysis of the variables used in the regression of

investment. The correlation between bank debt and investment is –0.093, and the

correlation between non-bank private debt and investment is -0.089. Both are statistically

significant at the 1% level. Investment and unused lines of credit have a coefficient of

0.080, statistically significant at the 5% level. Investment also has a positive correlation

with a firm’s cash flows, market-to-book ratio, and lagged capital expenditure and a

negative correlation with the leverage ratio.


Table 11 shows that borrowing from banks increases a firm’s investment while

borrowing from non-bank financial intermediaries decreases a firm’s investment. In the

first three regressions, the coefficients on outstanding bank debt and non-bank private debt

are negative and the coefficients on unused lines of credit are positive. However, when

instrumental variables are used to control for the endogeneity problem, the coefficients on

outstanding bank loans and unused lines of credit are positive and the coefficients on non-

bank private debt are negative. For example, in regression 5, the coefficient on non-bank

private debt is –0.442 and the coefficient on all bank private debt is 0.652, both statistically

significant at the 10% level. A firm’s investment also increases significantly with its last-

period capital expenditure. These findings suggest that bank borrowing reduces

information asymmetry and increases investment while non-bank private borrowing

constrains asset substitution and decreases investment.

Results in the regressions of cash holdings, equity risk, and investment altogether

suggest that banks are unique in reducing information asymmetries. The use of both

outstanding bank loans and unused lines of credit increases investment through the

reduction of information asymmetry and decreases a firm’s equity risk through monitoring.

Unused lines of credit also decrease a firm’s cash holdings by reducing information

30

asymmetry. But the use of non-bank private debt only decreases a firm’s equity risk and

investment through monitoring.

Results are also robust in a simultaneous system of equations. Financial decisions on

cash holdings, equity risk, and investment can be interrelated. Running regressions of cash

holdings, equity risk, and investment separately cannot capture such interrelations.

Previous findings still hold when I run the regressions of cash holdings, equity risk,

investment, leverage, outstanding bank debt, non-bank private debt, and unused lines of

credit in a simultaneous system of equations.

VI. Conclusion

This paper shows that not only the amount of debt, but also the sources of debt matter in

a firm’s financial decisions. I conduct several independent sets of tests to show the impacts

of outstanding bank loans, loans from non-bank financial intermediaries, and unused bank

lines of credit on a firm’s cash holdings, equity risk, and investment. In the context of

above tests, this paper shows the differences between private debt and public debt, bank

loans and loans from non-bank financial intermediaries, and outstanding private borrowing

and unused lines of credit.

Different types of private debt have distinctive effects on a firm’s cash holdings, equity

risk, and investment. Firms with more outstanding bank loans have a lower equity risk and

more cash holdings and investment. Firms with more non-bank private debt have a lower

equity risk and less investment. Moreover, firms with more unused bank lines of credit

have lower cash holding and equity risk but more investment.

In order to investigate the effects of private borrowing on a firm’s financial decisions, I

also examine the determinants of private borrowing to deal with the endogeneity problem

of private financing. I find that the use of bank loans and non-bank private debt decreases

with a firm’s size, age, market-to-book ratio, and fixed-asset ratio and increases with the

residual standard deviation. The amount of unused bank lines of credit decreases with a

firm’s size, age, fixed-asset ratio, and residual standard deviation and increases with the

31

market-to-book ratio. Moreover, I find that as firms grow bigger and older, they are more

likely to have a line of credit with banks for their investment opportunities than have

outstanding bank loans.

The results suggest that financial intermediaries have comparative advantages in

collecting information and monitoring over public investors and the uniqueness of private

financing affects a firm’s financial decisions. The use of private debt can mitigate one or

both of the adverse selection and asset substitution problems. In addition, different types of

private debt are different. Banks are better at reducing information asymmetry than non-

bank financial intermediaries. Unused line of credit is the only type of private debt that

serves as a substitute for cash.

32

Reference:

Aoki, Masahiko, 1994, “Monitoring characteristics of the main bank system: an

analytical and developmental view.” In Masahiko Aoki and Hugh Patrick, The Japanese

Main Bank System. New York: The Oxford University Press

Barclay, M., and Smith, C., 1995, The maturity structure of debt, The Journal of Finance

50, 609-631

Berlin, Mitchell and Jan Loeys, 1988, Bond Covenants and Delegated Monitoring, The

Journal of Finance, Volume XLIII, Issue 2.

Bhagat, S., M. Marr, and R. Thompson, 1985, The rule 415 experiment: equity markets,

The Journal of Finance 40, 1385-1401

Blackwell, D., and Kidwell, D., 1988, An investigation of cost differences between public

sales and private placements of debt, Journal of Financial Economics 22, 253-278

Blackwell, David W., M. Marr, and M. Spivey, 1990, Shelf registration and the reduced

due diligence argument: implications of the underwriter certification and the implicit

insurance hypotheses, Journal of Financial and Quantitative Analysis 25, 245-259

Boyd, John H. and Edward C. Prescott, 1986, Financial intermediary-coalitions, Journal of

Economic Theory 38, 211-232

Campbell, Tim S. and William A. Kracaw, 1980, Information production, market

signaling, and the theory of financial intermediation, The Journal of Finance, Volume

XXXV, Issue 4

Carey, M., Prowse, S., Rhea, J., and Udell, G., 1993, The economics of the private

placement markets: a new look, Financial markets, Institutions, and Instrument 2, 1-66

Diamond, Douglas W. 1984, Financial intermediation and delegated monitoring, The

Review of Economic Studies, Volume 51, Issue 3 (Jul. 1984), P393-414

Diamond, Douglas W. 1989, Reputation acquisition in debt markets, The Journal of

Political Economy, Volume 97, Issue 4 (Aug. 1989), P828-862

Dierkens, N., 1991, Information asymmetry and equity issues, Journal of Financial and

Quantitative Analysis 26, 181-199

33

Fama, Eugene F., 1985, What’s different about banks?, Journal of Monetary Economics

15, 29-39

Fazzari, Steve, Glenn Hubbard, and Bruce Petersen, 1988, Finance constraints and

corporate investment, Brookings papers on economic activity 1, 141-195

Galai, D., and Masulis, R., 1976, The option pricing model and the risk factor of stock,

Journal of Financial Economics 3, 53-81

Graham, J. and C. R. Harvey, 2001, The theory and practice of corporate finance: evidence

from the field, Journal of Financial Economics 60, 187-243

Hoshi, Takeo, Anyl Kashyap, and David Scharfstein, 1991, Corporate structure, liquidity,

and investment: Evidence from Japanese panel data, Quarterly Journal of Economics

106, 33-60

James, Christopher, 1987, Some evidence on the uniqueness of bank loans, Journal of

Financial Economics 19

James, Christopher and Peggy Wier, 1990, Borrowing relationships, intermediation, and

the cost of issuing public securities, Journal of Financial Economics 28, 149-171

Jensen, Michael C., and William H. Meckling, 1976, Theory of the firm: Managerial

behavior, agency costs and ownership structure, Journal of Financial Economics 3, 305-

360

Krishnaswami, Sudha, Paul A. Spindt, and Venkat Subramaniam, 1999, Information

asymmetry, monitoring, and the placement structure of corporate debt, Journal of

Financial Economics, Volume 51, 407-434

Lang, Larry, Eli Ofek and Rene M. Stulz, 1996, Leverage, investment, and firm growth,


Leland, Hayne E. and David H. Pyle, 1977, Informational asymmetries, financial structure,

and financial intermediation, The Journal of Finance, Volume 32, Issue 2, 371-387

Liu, Yang, 2004, Credit ratings and the use of private debt, Working paper, University of

Washington

Lummer, Scott L. and John J. McConnell, 1989, Further evidence on the bank lending

process and the capital market response to bank loan agreements, Journal of Financial

34

Economic 25

Mikkelson, Wayne H. and M. Megan Partch, 1985, Valuation effects of security

offerings and the issuance process, Journal of Financial Economics 15

Myers, Stewart C., 1977, Determinants of corporate borrowing, Journal of Financial

Economics 5, 147-175

Myers, Stewart C. 1984, The capital structure puzzle, The Journal of Finance, Vol. 39,

Issue 3

Myers, Stewart C. and Nicholas S. Majluf, 1984, Corporate financing and investment

decisions when firms have information that investors do not have, Journal of Financial

Economics 13

Nakamura, L., 1993, Commercial bank information: Implications for the structure of

banking, Structural Change in Banking, M.Klausner and L. White, eds. Homewood, IL:

Business One/Irwin

Opler, T., L. Pinkowitz, R. Stulz, and R. Williamson, 1999, The determinants and

implications of corporate cash holdings, Journal of Financial Economics 52, 3-46

Ramakrishnan, Ram T.S. and Anjan V. Thakor, 1984, Information reliability and a theory

of financial intermediation, The Review of Economic Studies, Volume 51, Issue 3

Rajan, R. G., 1992, Insiders and outsiders: The choice between informed and arm’s

length debt, Journal of Finance 47, 1367-1400

Smith, C., 1986, Investment banking and the capital acquisition process, Journal of

Financial Economics 15, 3-29

Smith, C., and Warner, J., 1979, On financial contracting: an analysis of bond covenants,


Table 1: Summary statistics

The sample period is 1996-2000. Included firms are manufacturing firms (SIC 2000-3999) who are trading in NYSE and AMEX. All data are obtained from Compustat, CRSP, and Moody’s industrial manuals. There are 961 unique firms and 3,399 firm-year observations. Cash holding is cash and other marketable securities divided by total assets. Equity risk is the standard deviation of daily stock returns. Investment is net capital expenditures for year +1 divided by the book value of fixed assets at the end of year 0. All private debt is the ratio of the sum of bank loans and non-bank private debt to total long-term debt. Bank debt is the ratio of long-term bank debt to total long-term debt. Non-bank private debt is the ratio of other long-term privately placed debt to total long-term debt. Unused line of credit is scaled by total assets. Market-to-book is the ratio of the market value of assets to the book value of assets. Size is measured by log of sales, which are stated in 1996 dollars. Leverage is the ratio of the book value of debt to the book value of total assets. Cash flow is earnings before interest expense and depreciation, scaled by total assets. R&D is research and development expense divided by total assets. Net working capital is working capital net of cash, scaled by total assets. Fixed-asset ratio equals the ratio of fixed assets to total assets. Residual standard deviation is defined as the standard deviation of the residuals of the market model regression using daily returns from the previous year. Age is the number of years since a firm’s incorporation. Cash holdings, all private debt, bank debt, non-bank private debt, market-to-book, leverage, cash flow, R&D, net working capital, fixed-asset ratio, and residual standard deviation are measured at the beginning of each fiscal year.

Variable Mean Std. Dev. 25% quartile Median 75% quartile Observation Cash holdings 0.087 0.125 0.013 0.037 0.110 3399 Equity risk 0.026 0.013 0.018 0.023 0.031 2587 Investment 0.041 0.219 -0.033 0.024 0.094 3399 Bank debt 0.222 0.329 0.000 0.041 0.393 3399 Non-bank private debt 0.070 0.210 0.000 0.000 0.000 3399 All private debt 0.293 0.370 0.000 0.046 0.593 3399 Unused lines of credit 0.071 0.110 0.000 0.000 0.121 3399 Market-to-book 1.842 1.141 1.174 1.495 2.094 3369 Size 6.433 1.865 5.338 6.478 7.661 3392 Leverage 0.247 0.172 0.117 0.238 0.351 3394 Cash flow 0.117 0.094 0.090 0.122 0.157 3304 R&D 0.028 0.049 0.000 0.013 0.036 3399 Net working capital 0.154 0.154 0.049 0.142 0.251 3332 Fixed-asset ratio 0.302 0.165 0.179 0.275 0.396 3399 Residual standard deviation 0.025 0.014 0.017 0.022 0.029 2605 Age 37.719 31.579 11.000 28.000 62.000 3388

Table 2: Descriptive statistics of unused bank lines of credit

The sample period is 1996-2000. Included firms are manufacturing firms (SIC 2000-3999) who are trading in NYSE and AMEX. All data are obtained from Compustat, CRSP, and Moody’s industrial manuals. There are 961 unique firms and 3,399 firm-year observations. 406 sample firms have data in all 5 years. Unused line of credit is scaled by total assets.

Firm Category N Unused bank lines of credit Mean

Firms do not have unused bank lines of credit 62 0

Firms have unused bank lines of credit in one year

64 0.027

Firms have unused bank lines of credit in two years

81 0.061

Firms have unused bank lines of credit in three years

84 0.096

Firms have unused bank lines of credit in four years

73 0.115

Firms have unused bank lines of credit in five years 42 0.052

Total 406 0.071

Table 3 Holdings of different types of private debt

The sample period is 1996-2000. Included firms are manufacturing firms (SIC 2000-3999) who are trading in NYSE and AMEX. All data are obtained from Compustat, CRSP, and Moody’s industrial manuals. There are 961 unique firms and 3,399 firm-year observations. Bank debt is the ratio of long-term bank debt to total assets. Non-bank private debt is the ratio of other long-term privately placed debt to total assets. Unused line of credit is scaled by total assets.

Firm Category N Bank debt

Mean

Non-bank private debt

Mean

Unused lines of credit Mean

Firms have outstanding bank loans only 88 0.144 Firms have outstanding non-bank private debt only 28 0.078 Firms have unused lines of credit only 136 0.090 Firms have outstanding bank loans and non-bank private debt only

35 0.070 0.095

Firms have outstanding bank loans and unused lines of credit only

386 0.094 0.096

Firms have outstanding non-bank private debt and unused lines of credit only

33 0.061 0.076

Firms have all three types of private debt 142 0.080 0.064 0.085 Firms with private borrowing relationships 848 0.071 0.021 0.079 Firms without private borrowing relationships 113 Total 961 0.064 0.019 0.071

Table 4 The use of different types of private debt The sample period is 1996-2000. Included firms are a sample of manufacturing firms (SIC 2000-3999)

who are trading in NYSE and AMEX. All data are obtained from Compustat, CRSP and Moody’s industrial manuals. Bank debt is the ratio of bank debt to total assets. Non-bank private debt is the ratio of other long-term privately placed debt to total assets. Unused lines of credit are scaled by total assets. Size is the natural logarithm of sales. Age is the number of years since a firm’s incorporation. Market-to-book is the ratio of the market value of assets to the book value of assets. Fixed-asset ratio equals the ratio of the value of fixed assets to the value of total assets. Residual standard deviation is defined as the standard deviation of the residuals of the market model regression using daily returns from the previous year. All variables are measured at the firm-level mean. P-values associated with Chi2 statistics are in the parentheses.

Dependent Variable Bank debt Non-bank private debt Unused lines of creditIntercept -0.791b -1.501b -1.492c (0.049) (0.013) (0.000) Size -0.130c -0.120a -0.045b (0.001) (0.055) (0.033) Ln(Age+1) -0.202c -0.122a -0.121c (0.000) (0.090) (0.001) Market-to-book -0.083 -0.031 0.104c (0.172) (0.582) (0.004) Fixed-asset ratio -0.326 -0.239 -0.175 (0.284) (0.627) (0.367) Residual standard deviation 4.869a 6.277 -9.017c (0.321) (0.361) (0.003) -log likelihood 1050.955 379.867 884.054

a Significant at the 0.10 level. b Significant at the 0.05 level. c Significant at the 0.01 level.

Table 5 Differences in the use of different types of private debt The sample period is 1996-2000. Included firms are a sample of manufacturing firms (SIC

2000-3999) who are trading in NYSE and AMEX. All data are obtained from Compustat, CRSP and Moody’s industrial manuals. Outstanding private debt is the sum of the amount of outstanding bank loans and loans from non-bank financial intermediaries. Outstanding and unused borrowing from financial intermediaries is the sum of outstanding private debt and unused lines of credit. Size is the natural logarithm of sales. Age is the number of years since a firm’s incorporation. Market-to-book is the ratio of the market value of assets to the book value of assets. Fixed-asset ratio equals the ratio of the value of fixed assets to the value of total assets. Residual standard deviation is defined as the standard deviation of the residuals of the market model regression using daily returns from the previous year. All variables are measured at the firm-level mean. P-values associated with Chi2 statistics are in the parentheses.

Dependent Variable Bank loans /Outstanding private debt

Unused lines of credit/ outstanding and unused borrowing from

financial intermediaries Intercept -0.073 -1.024c (0.262) (0.000) Size 0.003 0.043c (0.575) (0.002) Ln(Age+1) -0.005 0.040a (0.563) (0.054) Market-to-book 0.004 0.085c (0.733) (0.001) Fixed-asset ratio 0.006 -0.067 (0.920) (0.618) Residual standard deviation 0.209 -1.083 (0.901) (0.622) -Log likelihood 142.721 699.030


Table 6 Correlation analysis of the determinants of cash holdings The sample period is 1996-2000. Included firms are manufacturing firms (SIC 2000-3999) who are trading in NYSE and AMEX. All data are

obtained from Compustat, CRSP, and Moody’s industrial manuals. Cash is a firm’s cash holdings divided by its total assets. Bank debt is the ratio of long-term bank debt to total long-term debt. Non-bank private debt is the ratio of other long-term privately placed debt to total long-term debt. Market-to-book is the ratio of the market value of assets to the book value of assets. Size is measured by the natural logarithm of sales. Leverage is the ratio of the book value of debt to the book value of total assets. Cash flow is earnings before interest and depreciation, scaled by total assets. R&D is the research and development expense divided by total assets. Net working capital is working capital net of cash, scaled by total assets. Each variable is measured at its firm-level time-series mean. P-values are in parentheses.

Cash

Bank debt

Non-bank Private debt

Unused lines of credit

Market -to-book Size Leverage Cash flow R&D

Bank debt -0.271c (0.000) Non-bank private debt -0.075b -0.059a (0.023) (0.073)Unused lines of credit -0.140c 0.048 -0.038 (0.000) (0.146) (0.256)Market-to-book 0.279c -0.117c -0.012 -0.005 (0.000) (0.000) (0.723) (0.872) Size -0.427c -0.095c -0.092c 0.067b 0.001

(0.000) (0.004) (0.005) (0.043) (0.983) Leverage -0.363c 0.484c 0.200c 0.011 -0.108c 0.128c

(0.000) (0.000) (0.000) (0.743) (0.001) (0.000) Cash flow -0.252c 0.039 -0.017 0.108c -0.025 0.275c 0.013

(0.000) (0.235) (0.628) (0.001) (0.449) (0.000) (0.702)R&D 0.447c -0.129c 0.091c -0.057a 0.462c -0.257c -0.135c -0.318c

(0.000) (0.000) (0.006) (0.086) (0.000) (0.000) (0.000) (0.000)Net working capital -0.129c 0.018 0.073b 0.028 -0.227c -0.227c -0.222c 0.097c -0.138c (0.000) (0.582) (0.027) (0.396) (0.000) (0.000) (0.000) (0.003) (0.000)


41

Table 7 Cash holdings and unused bank lines of credit

The sample period is 1996-2000. Included firms are manufacturing firms (SIC 2000-3999) who are trading in NYSE and AMEX. All data are obtained from Compustat, CRSP, and Moody’s industrial manuals. Cash is a firm’s cash holdings divided by its total assets. Bank debt is the ratio of long-term bank debt to total assets. Non-bank private debt is the ratio of other long-term privately placed debt to total assets. Unused line of credit is the ratio of unused bank line of credit to total assets. All bank debt is the sum of bank debt and unused line of credit. All outstanding private debt is the sum of bank debt and non-bank private debt. Market-to-book is the ratio of the market value of assets to the book value of assets. Size is measured by ln(sales). Leverage is the ratio of the book value of debt to the book value of total assets. Cash flow is earnings before interest and depreciation, scaled by total assets. R&D is the research and development expense divided by total assets. Net working capital is working capital net of cash, scaled by total assets. All variables are measured at the beginning of each fiscal year. Each variable is measured at its firm-level time-series mean.

Ln(age+1), the fixed-asset ratio, size2, and the residual standard deviation are used as instrumental variables, when endogeneity is controlled for. Age is the number of years since a firm’s incorporation. The fixed-asset ratio equals the ratio of fixed assets to total assets. The residual standard deviation is defined as the standard deviation of the residuals of the market model regression using daily returns from the previous year. Coefficients are estimated using generalized method of moments (GMM). P-values reported in parentheses are corrected for heteroskedasticity.

42

Table 7 Cont’d

Independent variable Cash (1) Cash (2) Cash (3) Cash (4) Cash (5) Cash (6) Intercept 0.325c 0.325c 0.325c 0.615c 0.376c 0.396c (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) Bank debt -0.168c 0.460c (0.000) (0.000) Non-bank private debt -0.160a

-0.154a -0.261 -0.791 (0.063) (0.074) (0.312) (0.515) Unused line of credit -0.121c -0.122c -0.490c -0.618a (0.001) (0.001) (0.000) (0.094) All bank debt -0.148c -0.361b (0.000) (0.024) All outstanding private debt

-0.166c 0.310a (0.000) (0.073)

Market-to-book ratio 0.010b 0.010b 0.010b 0.046c 0.006 0.010b (0.031) (0.030) (0.030) (0.000) (0.214) (0.044) Size -0.027c -0.027c -0.027c -0.014c -0.021c -0.028c (0.000) (0.000) (0.000) (0.005) (0.003) (0.000) Leverage -0.166c -0.174c -0.174c -0.479c -0.447c -0.225c (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) Cash flow -0.017 -0.017 -0.017 -0.001 -0.017 -0.012 (0.504) (0.511) (0.505) (0.934) (0.425) (0.581) R&D 0.428c 0.429c 0.429c 0.302a 0.392b 0.337b (0.000) (0.000) (0.000) (0.059) (0.016) (0.013) Net working capital -0.195c -0.195c -0.194c -0.176c -0.172c -0.206c (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Adjusted R2 0.473 0.473 0.474 0.341 0.341 0.340N 922 922 922 893 893 893Instrumental variable Yes Yes Yes a Significant at the 0.10 level. b Significant at the 0.05 level. c Significant at the 0.01 level.

43

Table 8 Correlation analysis of the determinants of equity risk

The sample period is 1996-2000. Included firms are manufacturing firms (SIC 2000-3999) who are trading in NYSE and AMEX. All data are obtained from Compustat, CRSP, and Moody’s industrial manuals. Equity risk is the standard deviation of daily stock returns. Bank debt is the ratio of long-term bank debt to total long-term debt. Non-bank private debt is the ratio of other long-term privately placed debt to total long-term debt. Size is measured by the ln(sales). Leverage is the ratio of the book value of debt to the book value of total assets. Bank debt, non-bank private debt, all private debt, and leverage are measured at the beginning of each fiscal year. Each variable is measured at its firm-level time-series mean. P-values are in parentheses.

Equity risk Bank debt Non-bank private debt

Unused lines of credit

Size

Bank debt 0.101c (0.002) Non-bank private debt 0.082b -0.058a (0.013) (0.077) Unused lines of credit -0.134c 0.058a -0.040 (0.000) (0.076) (0.224) Size -0.539c -0.096c -0.096c 0.061a (0.000) (0.003) (0.003) (0.062) Leverage 0.067b 0.476c 0.205c 0.021 0.151c (0.040) (0.000) (0.000) (0.525) (0.000)


Table 9 Equity risk and unused bank lines of credit The sample period is 1996-2000. Included firms are manufacturing firms (SIC 2000-3999) who are trading in NYSE and AMEX. All data are

obtained from Compustat, CRSP, and Moody’s industrial manuals. Equity risk is the standard deviation of daily stock returns. Bank debt is the ratio of long-term bank debt to total assets. Non-bank private debt is the ratio of other long-term privately placed debt to total assets. Unused line of credit is the ratio of unused bank line of credit to total assets. All bank debt is the sum of bank debt and unused line of credit. All outstanding private debt is the sum of bank debt and non-bank private debt. Size is measured by the ln(sales). Leverage is the ratio of the book value of debt to the book value of total assets. Bank debt, non-bank private debt, all private debt, and leverage are measured at the beginning of each fiscal year. Each variable is measured at its firm-level time-series mean.

The fixed-asset ratio, the residual standard deviation, size2, and the market-to-book ratio are used as instrumental variables, when endogeneity is controlled for. The fixed-asset ratio is the ratio of the book value of fixed assets to the book value of total assets. The residual standard deviation is defined as the standard deviation of the residuals of the market model regression using daily returns from the previous year. The market-to-book ratio is the ratio of the market value of assets to the book value of assets.

Coefficients are estimated using generalized method of moments (GMM). P-values reported in parentheses are corrected for heteroskedasticity.

Table 9 Cont’d

Risk (1) Risk (2) Risk (3) Risk (4) Risk (5) Risk (6) Intercept 0.051c 0.051c 0.051c 0.022c 0.025c 0.051c (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Bank debt

-0.004 -0.320c (0.420)

(0.000)Non-bank private debt -0.003 -0.005 -0.324c -0.228c (0.618) (0.440) (0.000) (0.001) Unused line of credit

-0.017c -0.017c -0.225c -0.479c (0.000) (0.000) (0.000) (0.000)

All bank debt -0.010c -0.271c

(0.003) (0.000) All outstanding private debt

-0.035 -0.052c (0.398) (0.002)

Size -0.004c -0.004c -0.004c -0.003c -0.003c -0.003c (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Leverage 0.014c 0.016c 0.014c 0.142c 0.131c 0.109c (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Adjusted R2 0.321 0.318 0.322 0.258 0.252 0.250N 931 931 931 900 900 900Instrumental variable Yes Yes Yes


46

Table 10 Correlation analysis of the determinants of investment

The sample period is 1996-2000. Included firms are manufacturing firms (SIC 2000-3999) who are trading in NYSE and AMEX. All data are obtained from Compustat, CRSP, and Moody’s industrial manuals. Investment is capital expenditures minus depreciation for year +1 divided by the book value of fixed assets at the end of year 0. Bank debt is the ratio of long-term bank debt to total long-term debt. Non-bank private debt is the ratio of other long-term privately placed debt to total long-term debt. Cash flow is earnings before depreciation and interest expense. Market-to-book is the ratio of the market value of assets to the book value of assets. Leverage is the ratio of the book value of debt to the book value of total assets. Capital expenditure is the one-year lag of the ratio of capital expenditure to fixed assets. Bank debt, non-bank private debt, market-to-book, and leverage are measured at the beginning of each fiscal year. Each variable is measured at its firm-level time-series mean. P-values are in parentheses.

Investment

Bank debt Non-bank

private debt Unused lines

of credit

Cash flow

Market-to-

book

Leverage

Bank debt -0.093c (0.005) Non-bank private debt -0.089c -0.058a

(0.006) (0.076)Unused lines of credit 0.080b 0.055a -0.037 (0.014) (0.095) (0.264) Cash flow 0.073b 0.041 -0.018 0.111c

(0.025) (0.206) (0.586) (0.001)Market-to-book 0.124c -0.116c -0.012 -0.005 0.038

(0.000) (0.000) (0.734) (0.876) (0.246)Leverage -0.115c 0.471c 0.192c 0.008 0.022 -0.111c

(0.000) (0.000) (0.000) (0.801) (0.511) (0.000)Capital expenditure (0) 0.727c -0.037 -0.007 0.047 0.031 -0.155c -0.089c (0.000) (0.256) (0.830) (0.156) (0.339) (0.000) (0.007)


47

Table 11 Investment and unused bank lines of credit The sample period is 1996-2000. Included firms are manufacturing firms (SIC 2000-3999) who are trading in NYSE and AMEX. All data are obtained from

Compustat, CRSP, and Moody’s industrial manuals. Investment is capital expenditures minus depreciation for year +1 divided by the book value of fixed assets at the end of year 0. Bank debt is the ratio of long-term bank debt to total assets. Non-bank private debt is the ratio of other long-term privately placed debt to total assets. Unused line of credit is the ratio of unused bank line of credit to total assets. All bank debt is the sum of bank debt and unused line of credit. All outstanding private debt is the sum of bank debt and non-bank private debt. Cash flow is earnings before interest expense and depreciation. Market-to-book is the ratio of the market value of assets to the book value of assets. Capital expenditure is the one-year lag of the ratio of capital expenditure to fixed assets. Leverage is the ratio of the book value of debt to the book value of total assets. Bank debt, non-bank private debt, all private debt, market-to-book, and leverage are measured at the beginning of each fiscal year. Each variable is measured at its firm-level time-series mean.

Ln(sales), ln(age+1), the fixed-asset ratio, size2, and the residual standard deviation are used as instrumental variables, when endogeneity is controlled for. Age is the number of years since a firm’s incorporation. The fixed-asset ratio is the ratio of the book value of fixed assets to the book value of total assets. The residual standard deviation is defined as the standard deviation of the residuals of the market model regression using daily returns from the previous year.

Coefficients are estimated using generalized method of moments (GMM). P-values reported in parentheses are corrected for heteroskedasticity.

48

Table 11 Cont’d

Investment (1) Investment (2) Investment (3) Investment (4) Investment (5) Investment (6) Intercept 0.006 0.014 0.006 -0.092b -0.019 -0.104b (0.696) (0.394) (0.723) (0.045) (0.747) (0.017) Bank debt -0.120b 0.521

(0.019) (0.130) Non-bank private debt

-0.257b -0.238b -0.126 -0.442a

(0.011) (0.020) (0.873) (0.059) Unused line of credit 0.094b 0.099b 0.599b 0.884c (0.023) (0.015) (0.018) (0.000) All bank debt -0.031 0.652a

(0.330) (0.070)All outstanding private debt -0.150c -0.507c (0.002) (0.008) Cash flow 0.043 0.048 0.044 0.038 0.039 0.028 (0.455) (0.419) (0.448) (0.425) (0.493) (0.465) Market-to-book 0.001 0.002 0.001 0.001 0.001 0.004 (0.897) (0.810) (0.914) (0.930) (0.960) (0.571) Capital expenditure (0) 0.566c 0.568c 0.566c 0.595c 0.608c 0.547c (0.000) (0.000) (0.000) (0.000) (0.000) (0.000) Leverage -0.001 -0.029 -0.002 -0.077 -0.015 -0.114 (0.996) (0.452) (0.959) (0.558) (0.922) (0.365) Adjusted R2 0.541 0.537 0.540 0.448 0.457 0.452N 934 934 934 891 891 891Instrumental variable Yes Yes Yes


the sources of debt matter, too

Documents