The impact of the Sarbanes-Oxley act on private debt contracting.

Pae, Sangshin "Sam"

INTRODUCTION

The purpose of this paper is to investigate whether an association exists among monitoring, debt covenants, and the cost of debt and whether it has been affected by certain provisions of the Sarbanes-Oxley Act of 2002 (hereafter SOX). I also examine whether the increased monitoring that resulted from SOX differentially affected firms with certain characteristics, such as size or presence of growth options.

The agency costs of debt arise because of conflicts between shareholders and debtholders and are mainly due to asset substitution and underinvestment problems which occur after the debt contract is finalized. Jensen and Meckling (1976) posit that shareholders have incentives to invest in high variance projects, i.e., risky and high expected return projects, at the expense of debtholders (asset substitution). Alternatively, Myers (1977) argues that shareholders have incentive to underinvest in positive net present value (NPV) projects because the positive expected NPV fails to cover previously promised debt repayments (underinvestment). To mitigate shareholder-bondholder agency costs, bondholders use three devices--monitoring, debt covenants, and cost of debt--to protect themselves from managerial opportunism. Thus, an exogeneously mandated increase in monitoring should produce less reliance on the other two methods, debt covenants and cost of debt, for reducing agency costs.

In this paper, I examine whether the event of increased regulatory monitoring produced less reliance on debt covenants, the cost of debt or both. To accomplish my objective, I compare the number of debt covenants and the cost of debt for private debt contracts during the pre-SOX and post-SOX periods. Using data on private debt issued during the 1999-2005 time period (specifically a sample of 4,610 facilities), I empirically examine the association among monitoring, debt covenants, and the cost of debt. I focus on private debt because, by its nature, private debt is more likely to have restrictive covenants than public debt (Kwan and Carleton, 2004). Due to the large number of bondholders involved in a public debt issue, renegotiating a debt contract following a debt covenant violation can be costly and difficult. Thus, private debt contracts generally contain a greater number of debt covenants than public debt contracts making it easier to examine the association among monitoring, debt covenants, and cost of debt in private debt contracts.

In the analysis that follows, I rely on the definition of monitoring in Jensen and Meckling (1976) (Jensen and Meckling define monitoring as more than just measuring or observing the behavior of the agent. It includes efforts on the part of the principal to "control" the behavior of the agent through various activities. I define "monitoring" in terms of broad measure which includes all internal and external monitoring that affects agent's behavior to reduce his or her discretion) and the specific provisions in SOX that increase the monitoring of public companies, especially Title II "Auditor Independence," Title III "Corporate Responsibility," and Title IV "Enhanced Financial Disclosures."

Previous studies on debt contracts document only the association between monitoring and debt covenants (Black et al., 2004) or the association between debt covenants and the cost of debt (Beatty et al., 2002). Black et al. (2004) find a negative relationship between monitoring and the frequency of debt covenants. In particular, they document decreases in the use of debt covenants during the periods of increased monitoring. In addition, Beatty et al. (2002) find a negative relationship between the inclusion of debt covenants and the cost of debt. They argue that managers are willing to bear higher interest rates to retain accounting flexibility. I extend the work in these two studies by evaluating a broader set of variables that are used to mitigate agency costs between shareholders and debtholders. Specifically, I hypothesize that during the period of increased monitoring induced by SOX (i.e., post-SOX period), the number of debt covenants or the cost of debt will decrease in private debt contracts due to interaction effects among monitoring, debt covenants, and cost of debt. In addition, I hypothesize that small firms and growth firms will demonstrate a much greater impact from the implementation of SOX because the Act will reduce conflicts between shareholders and debtholders more for high information-asymmetry firms.

The empirical results are generally consistent with my hypotheses. I find a statistically significant decrease in the cost of debt after the implementation of SOX. However, I do not find evidence that the usage of debt covenants decreased during the post-SOX period. The results indicate that exogeneously mandated monitoring produces less reliance on the cost of debt but not on debt covenants. In addition, I find evidence that, on average, small firms and growth firms are more influenced by increased monitoring due to SOX. These results are consistent with my hypothesis that firms with higher information asymmetry were more influenced by SOX.

The results demonstrated in this paper make a number of contributions to our understanding of debt contracting. First, this paper examines the association among all three devices that are used to minimize the conflicts between shareholders and debtholders. Prior research has only found evidence of an association between monitoring and debt covenants (Black et al., 2004) and an association between debt covenants and the cost of debt (Beatty et al., 2002). This paper is the first to study the relationships among all three methods. Second, while most previous studies related to SOX focused on stock price reactions and corporate governance issues (Berger et al., 2004; Jain and Rezaee, 2004; Li et al., 2004; Zhang, 2005), this paper addresses the impact of the SOX on debt contracts. Finally, this paper provides a foundation for future work on the relations between monitoring, debt covenants, and the cost of debt.

The remainder of the paper is organized as follows. Section 2 develops the hypotheses and presents the empirical models used to investigate the relation between monitoring, debt covenants, and the cost of debt. Section 3 describes the sample, data sources, and variable measurements. Section 4 provides descriptive statistics. Section 5 presents correlation analysis and my primary results. It also examines a potential endogeneity problem and presents instrumental variable approach results. Finally, in section 6, I conclude.

HYPOTHESES DEVELOPMENT AND EMPIRICAL MODELS

Herein are developed each of the principal hypotheses based on the impact of the increased monitoring imposed by SOX and the regression models used. The first two hypotheses consider the association among the three most important components of debt contracts: monitoring, debt covenants, and the cost of debt. The next two hypotheses consider whether the increased monitoring due to SOX had different effects based on the size of the firm or its growth options.

Prior empirical work on debt contracts shows that when the level of regulatory monitoring increases, banks in their role as borrowers reduce their use of debt covenants intended to reduce agency costs (Black et al., 2004). They argue that regulatory monitoring and debt covenants both strive to limit a bank's default risk, and since regulatory monitoring cannot be controlled by bank shareholders, they try to minimize the agency costs by substituting monitoring through debt covenants where debt covenants and regulatory monitoring intersect. SOX's provisions force increased levels of inside and outside monitoring on firms. For example, Section 202 requires that all auditing services and all permitted non-auditing services to be pre-approved by the client company's independent audit committee, Section 302 requires each public company's CEO and CFO to certify that they have reviewed the quarterly and annual reports their companies file with the SEC, and Section 403 requires most transactions by insiders to be electronically filed with the SEC within two business days. In the presence of this additional monitoring, lenders and borrowers may be able to reduce the number and type of covenants employed. I examine whether this substitution effect between monitoring and debt covenants can be generalized by expanding the sample of firms to include other industries besides banks, as examined in Black et al. (2004).

In addition, I develop a second hypothesis based on prior research in managerial opportunism. Jensen and Meckling (1976) suggest that covenants are included in debt contracts as a strategy for restricting managerial opportunism. They argue that by agreeing to restrict future opportunistic behavior, borrowers can reduce their current borrowing costs. Thus, the borrower faces a trade-off between retaining the possibility of future opportunistic behavior and obtaining a lower interest rate. Beatty et al. (2002) find evidence that borrowers are willing to pay substantially higher interest rates to retain accounting flexibility that may help them avoid covenant violations.

Black et al. (2004) only investigate the association between monitoring and debt covenants. While prior study (Beatty et al., 2002) suggests that the cost of debt and debt covenants could be substitutes, I hypothesize that there is also a substitution effect between monitoring and the cost of debt. For example, if the cost of debt increases and the number of debt covenants decreases at the same time, we cannot attribute the decreased number of debt covenants solely to the effect of increased monitoring. (Like most previous studies on debt covenants, I do not measure the tightness of debt covenants due to the cost of accessing actual debt covenant information. Prior studies use proxies such as debt-equity ratio (Duke and Hunt, 1990; Press and Weintrop, 1990; DeFond and Jiambalvo, 1991), direct measurements of covenant slack (Dichev and Skinner, 2002), or number of covenants (Black et al., 2004; Begley and Feltham, 1999) for measuring the tightness of debt covenants. Begley and Feltham (1999) find evidence that the existence and tightness of the covenants are highly positively correlated.) Thus, I expect increased monitoring to affect debt covenants, the cost of debt, or both. Because of its increased monitoring requirements, SOX should have altered the use of debt covenants and the cost of debt in post-SOX debt contracts. This leads to my first and second hypotheses (in alternative form):

Hypothesis 1: The number of debt covenants is likely to decrease, ceteris paribus, during the post-SOX period.

Hypothesis 2: The cost of debt is likely to decrease, ceteris paribus, during the post-SOX period.

I test these hypotheses by estimating (1) a logistic regression model and (2) a multiple regression model in which either debt covenants or spread is the dependent variable. Ten covariates are included to control for other potentially relevant explanatory factors, discussed below. The model is as follows:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (1)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (2)

where i refers to the facility (facility is a tranche of a private debt offer) and t refers to the time of issuance. Variable definitions are as follows:

COV = dummy variable set equal to one if the debt covenant is included in debt contract, zero otherwise.

SOX = an indicator variable, which is a proxy for monitoring set equal to one if the sample period is in the post-SOX period; zero if the sample period is in the pre-SOX period.

SIZE = natural log of total assets of firm at quarter-end.

LEV = ratio of long-term debt to total assets at quarter-end.

VAR = 5 years earnings variability prior to the debt contract; computed as the standard deviation of firm i's net income before extraordinary items (scaled by total assets) measured over rolling five year windows.

GROWTH = market-to-book ratio; computed as market value of equity divided by total book value of equity.

AMOUNT = total amount of money (scaled by total assets) that the firm borrowed in debt contract.

MAT = stated maturity calculated in years.

RATING = Moody's senior debt ratings information on each loan facility; ratings of Aaa, Aa, A, Baa, Ba, B, and lower than B represents 1 through 7, respectively.

COL = dummy variable set equal to one if the loan is secured; zero otherwise.

SPREAD = basis point spread over LIBOR inclusive of all fees, which is a proxy for cost of debt. In general, this spread is fixed over the life of the loan.

TBOND = 5 year Treasury bond rate.

The dummy variable, denoted SOX, is set equal to one for the post-SOX period and zero for the pre-SOX period to determine whether there were any changes in debt contracts before and after the implementation of the Act. I expect that this SOX variable, which is the main variable of interest, would be significantly negative due to the additional monitoring requirements associated with SOX. As a control variable, I use natural log of total assets, denoted as SIZE, to proxy for the size of the firm. Small firms are generally viewed as more risky to creditors, so I expect this SIZE variable to have a negative relationship with the dependent variables. In addition, large firms may be better able to manage risk, so firm size may affect interest rate choice.

Interest rates on loans are extremely sensitive to default risk. As a proxy for the loan's credit risk, I use Moody's senior debt ratings information on each loan facility, as provided in the Dealscan database, denoted by RATING. I code RATING into 1 through 7 to represent the ratings of Aaa, Aa, A, Baa, Ba, B and lower than B. Higher default risk (a higher value of debt rating) is related to higher interest rate and more debt covenants; therefore, I expect a positive relation between RATING and dependent variables.

Debt contracts are also affected by the leverage of the firm. Capital Structure Theory suggests that at relatively low debt levels, the probability of bankruptcy and financial distress is low and the benefit from debt outweighs the cost. However, as the debt level increases, the possibility of financial distress also increases, so the benefit from debt financing may be more than offset by the financial distress costs. Thus, the interest rate does not have a linear relationship with the firm's leverage, making it hard to predict whether leverage will have a positive or negative relationship with dependent variables. This difficulty is compounded because different firms and industries have different optimal levels of leverage. The variable LEV is the ratio of long-term debt over total capital (debt plus equity), and I do not predict a sign for [[alpha].sub.3] or [[beta].sub.3].

I use the standard deviation of the firm's return on assets (scaled by total assets) over the past five years, denoted by VAR, rather than the stock price variations to measure the variability of the firm's performance. Because debtholders usually look at the firm's earnings stability rather than the stock price movements, earnings variability is assumed to be a better proxy for measuring risk. Therefore, I expect VAR to have a positive relationship with dependent variables. Other control variables are debt characteristic variables such as the cost of debt, total amount of debt, maturity and whether the loan has collateral. Since debt covenants and cost of debt are substitutes (that is, when debt covenants increase, interest rates decrease and vice versa), I expect SPREAD to have a negative relationship with COV. The debt contract is also affected by its maturity, the total amount of the loan and whether it is secured by collateral. I expect MAT to have a positive relationship with dependent variables because longer maturity debt is riskier farther in the future because of increased uncertainty of repayment. Collateral allows the lender to recover, at least partially, the principal. When the borrower fails to make a promised payment, the lender can sell the collateral, thus reducing the likelihood or amount of loss on debt. Therefore, I expect a negative relationship between COL and SPREAD because if the debt contract is secured by collateral, the cost of debt should be lower. However, a number of prior studies have found a positive relation between the existence of collateral and the cost of debt (Berger and Udell, 1990; John et al., 2003). Berger and Udell (1990) and John et al. (2003) suggest that lower quality firms are required to use collateral when issuing debt, while higher quality firms are able to issue without it.

In addition, a firm's growth opportunities are reflected in the market-to-book ratio. Growth firms have more intangible assets whose valuation depends heavily on future profitability. Therefore, I expect higher market-to-book ratio (i.e., growth firms), denoted by GROWTH, to have a positive relationship with SPREAD. To control for the effects of extreme values, I remove those observations that have negative or zero book value.

During my sample period, economic environments were unstable, and to capture the macroeconomic factors, I obtained five-year Treasury-bond rates from the United States Department of the Treasury (http://www.treasury.gov).

Since SOX contains substantive reforms with respect to financial reporting, we would expect it to have a significant impact on reducing information asymmetry. In particular, firms that significantly manage earnings should be impacted differently than firms that do not manage earnings. In the bond market, as in the stock market, the risk premium is different depending on the level of information asymmetry. The level of information asymmetry for small firms is more likely to be reduced than for large firms as a result of SOX. Thus, the relative reduction in the risk premium on debt contracts is likely to be greater for smaller firms, and thus, I expect a greater reduction in the cost of debt issued during the post-SOX period for smaller firms.

As just noted, prior studies show that the level of information asymmetry differs between small firms and large firms. For example, Lakonishok and Lee (2001) find that insider purchases in smaller firms predict future returns, but this predictive power does not hold for larger firms. Similarly, Finnerty (1976) and Seyhun (1986) find insider profits are larger for smaller firms. If firm size is proxy for information asymmetry and information asymmetry is greater in smaller firms, then the results of Lakonishok and Lee (2001), Seyhun (1986), and Finnerty (1976) suggest insider profits are larger when greater information asymmetry is present. A recent paper by Bharath et al. (2006) also documents that small borrowers have greater information asymmetries. In addition, Dixon et al. (2006) posit that because small businesses are likely to be less diversified and less able to leverage economies of scale or to access capital markets, the cost of complying with a particular regulation may be different for smaller and larger firms. Thus, my third hypothesis is (in alternative form):

Hypothesis 3: Small firms are more likely to have a greater relative reduction in the cost of debt after the passage of SOX than larger firms.

Barclay and Smith (1995) suggest that a firm's future investment opportunities may be viewed as options whose value depends on the likelihood that the firm will exercise the options optimally. Therefore, the contracting costs due to underinvestment and asset substitution are higher for firms with more growth options because the conflict between shareholders and bondholders over the exercise of the options is greater. Shareholders of high-growth firms can more easily substitute riskier projects for less risky ones and are also more susceptible to foregoing positive NPV projects if the gains accrue predominantly to the bondholders. That is, management may invest in high-risk negative NPV projects that increase the value of the equity but decrease the value of the debt. Consistent with my third hypothesis, if we expect the Sarbanes-Oxley Act to have a significant impact on reducing information asymmetry, then it should have a greater impact on firms that have more growth options because the conflict between shareholders and bondholders is greater for such firms, and greater conflict means that the level of information asymmetry is greater for these firms. Thus, when the information asymmetry is reduced due to Sarbanes-Oxley, the risk premium of debt contracts will decrease. This leads to my fourth hypothesis (in alternative form):

Hypothesis 4: Firms with higher growth options are more likely to have a greater relative reduction in the cost of debt after the passage of SOX than firms with lower growth options.

I test my third and fourth hypotheses by extending the multiple regression model that I used to test the second hypothesis. I include two interaction terms to determine the impact of SOX on variables of interest (SIZE and GROWTH). The model is as follows:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (3)

SAMPLE SELECTION

The data for this analysis is drawn from three sources (Dealscan, Quarterly and Annual Compustat, and the United States Department of Treasury). I collect debt characteristic variables from Dealscan (See Dichev & Skinner, 2002 for a discussion of the Dealscan database), firm characteristic variables from Compustat, and a macroeconomic control variable from the United States Department of Treasury (http://www.treasury.gov). I first draw the initial sample of firms that issued debt during the period from January 1999 to December 2005. I define the three year period from January 1999 to December 2001 as the pre-SOX period and the three year period from January 2003 to December 2005 as the post-SOX period. Because of the news coverage, the market should have known about SOX in 2002, although not about the exact provisions of the Act, before it was actually signed by President Bush. Therefore, if the market participants were rational, we should assume that the debt contracts could have been affected by SOX during this period. Thus, I set the one-year period from January 2002 to December 2002 as a transition period.

Typically, loan deals are broken into individual facilities. Each of the facilities is a representation of a different tranche of the total loan. The facilities differ in terms of maturity and spread. They also differ in terms of debt covenants and structure. From Dealscan, only newly issued facilities and the details of the contract for each facility, such as amount of loan issued, interest rate (spread), date of issue, date of maturity, Moody's senior debt rating, collateral, and information about debt covenants, were collected during the sample period. I assigned a score of one if the debt contract had certain types of debt covenants (either the financial debt covenants or general debt covenants) and zero otherwise. In addition, I assigned a score of one if the facility had collateral and zero otherwise. The sample selection process is described in Table 1. Since Dealscan does not offer CUSIP numbers, ticker symbols are used for matching with Compustat. Annual Compustat was used to calculate earnings variability. I eliminate firms that had zero or negative book value of equity.

DESCRIPTIVE STATISTICS

Panel A of Table 2 reports the means, medians, and standard deviations for variables used in the multiple regressions. As shown in Panel A of Table 2, the sample of 4,610 facilities exhibits considerable variation in firm size and borrowing amount. There were total of 1,000 firms in my data sample. There were 792 firms that issued debt during the pre-SOX period and 938 firms that issued debt during the post-SOX period. Firm size during the sample period ranges from a minimum of $3 million to a maximum of $242 billion. Facility amount during the sample period ranges from a minimum of $0.1 million to a maximum of $25 billion. The sample debt issues are quite large. The median amount issued is $175 million. The facilities in my sample appear to be risky. The median credit rating of 7 indicates that more than half of the sample contracts are entered into by below-investment-grade borrowers. This is mainly because Dealscan consists of only private debt. A previous study on debt choice by Denis and Mihov (2003) shows that public borrowers are larger and have higher credit ratings than firms borrowing from either banks or non-bank private lenders. Their findings suggest that firms with the highest credit quality borrow from public sources, firms with medium credit quality borrow from banks, and firms with the lowest credit quality borrow from non-bank private lenders. Thus, my sample is actually downward biased in terms of size and credit quality because of the nature of the private debt market and the limitation of my database. In addition, slightly less than half of the facilities in my sample required collateral and more than half had some type of covenants, either financial covenants or general covenants.

To examine the differences between debt contracts for firms that borrowed during the pre-SOX period and those that borrowed during the post-SOX period, I first divide the sample into two groups: pre-SOX and post-SOX. Panel B of Table 2 shows the different characteristics between firms that borrowed during the pre-SOX and post-SOX periods by using two-sample t-tests. Panel B of Table 2 presents the differences in spread, size, leverage, earnings variability, market-to-book ratio, and other characteristics between the two sub-samples. Private debt issued during the post-SOX period typically has larger spread, is highly leveraged, has longer maturity, is less risky, and has more collateral and more debt covenants than the debt issued during the pre-SOX period. Specifically, maturity and debt covenants are significantly different while spread, leverage, Moody's rating, and collateral are only marginally significant. In addition, the 5-year T-bond rate for pre-SOX firms is significantly higher than the rate for post-SOX firms, which shows that there were large economic fluctuations during the sample period. The difference between 5-year T-bond rates shows that the economic environment has changed dramatically during these periods. During my sample period, the highest 5 year T-bond rate was 6.83% on May 8, 2000 and the lowest 5 year T-bond rate was 2.08% on June 13, 2003. However, Panel B of Table 2 shows that firm size, earnings variability, market-to-book ratio, and the amount of borrowing are not statistically different between the two sub-samples.

The mean spread for firms in the pre-SOX period is 169 basis points while the mean spread for firms in the post-SOX period is 177 basis points, marginally significantly higher at the 10% level. Also, firms in the pre-SOX period have an average of 3.05 years of duration, while the maturity for post-SOX firms is 3.78 years which is significantly longer.

RESULTS

Correlation Analysis

I present Pearson correlation coefficients for the pooled regression variables in Table 3. The correlation between SPREAD and SOX indicates that the cost of debt for debt contract has increased during the post-SOX period. The marginally significant change in SPREAD seen in Table 2, Panel B also indicates the increase in interest rates. The variable SIZE is significantly correlated with several variables. SIZE is positively correlated with LEV, AMOUNT, and RATING and negatively correlated with SPREAD, MAT, COL, COV, and TBOND. This positive relationship indicates that (at least in my sample) large firms are more leveraged and borrow more money per facility than small firms. The negative relationship indicates that the debt contracts for large firms have lower interest rates, shorter maturities, are more often unsecured, and have fewer restrictive debt covenants than those of small firms. The significantly positive correlation between GROWTH and SPREAD indicates that growth firms are viewed as more risky; thus the cost of debt is higher for growth firms. In addition, the significantly positive correlation between GROWTH and RATING also indicates that the growth firms have lower credit ratings, which is consistent with growth firms being more risky. The variable AMOUNT is significantly negatively correlated with SPREAD, SIZE, MAT, RATING, COL, and COV. This indicates that the firm that borrows a lot of money from a one-time deal generally has a low interest rate, has shorter maturity, has quite low credit risk, is unsecured, and has fewer debt covenants on the deal.

Table 3 also reveals that many explanatory variables are significantly correlated with each other. The formal hypotheses tests are based on logistic and multiple regression analysis.

Logistic and Multiple Regression Results

In Table 4, I report logit regression results for dichotomous dependent variable COV. Consistent with prior research (Beatty et al., 2002), I find evidence that debt covenants and the cost of debt have a negative relationship. However, the main variable of interest is SOX, a dichotomous variable set to one if the year is in the post-SOX period, and logit regression results in Table 4 show no evidence that the use of debt covenants decreased during the post-SOX period. I next turn to the results in Table 5 to determine whether the cost of debt decreased due to increased monitoring in the post-SOX period.

I present the estimation results of the multiple regressions for 4,610 firm-quarter observations in Table 5. The variable of interest in these regressions is again SOX. Model 1 in Table 5 reports the univariate analysis between SPREAD and SOX. Consistent with Panel B of Table 2 and Table 3, the univariate analysis between SPREAD and SOX indicates that the cost of debt has increased during the post-SOX period. I include all of the control variables in the Full Model in Table 5. The results of estimating the Full Model show that the sign of the coefficient on SOX is significantly negative, suggesting that greater monitoring is associated with a lower cost of debt capital after controlling for firm and debt characteristic variables.

Consistent with my second hypothesis, shareholders appear to be substituting a reduced cost of debt for increased monitoring in the post-SOX period. The main variable of interest, SOX, is significantly negative, and its magnitude suggests a decrease of 29 basis points in the firm's cost of debt during the post-SOX period, after controlling for other factors. The signs of the coefficients on the control variables are generally consistent with my predictions except the variable COL. The relationship between the cost of debt and collateral is opposite to my prediction, but consistent with other previous empirical studies (Berger and Udell, 1990; John et al., 2003). Berger and Udell (1990) and John et al. (2003) posit that the positive association between the cost of debt and collateral might arise because lenders require collateral on lower quality firms. Careful examination of the results in Tables 4 and 5 together shows that monitoring is a substitute for the cost of debt but not debt covenants. Hence, I conjecture that the lenders might be willing to maintain a certain minimum level of debt covenants. Beatty et al. (2002) posit that borrowers are willing to pay substantially higher interest rates to retain accounting flexibility. Since the number of debt covenants has not decreased in the post-SOX period, I conjecture that the result may be due to the lender's willingness to maintain debt covenants at a certain level.

The Extended Model in Table 5 shows the result of the extended multiple regression model, which I use to test hypotheses 3 and 4. I extend the previous multiple regression model by including two interaction terms, which are SOX*SIZE and SOX*GROWTH. If the increased post-SOX monitoring influenced the cost of debt, this regression should show what kind of firms are more (or less) affected by increased monitoring. The coefficient for SOX*SIZE is significantly positive. This means that in the post-SOX period (i.e., when SOX is equal to one) the magnitude of the coefficient for SIZE variable is reduced, which is evidence that the gap between small and large firms has decreased. In the pre-SOX period alone, the coefficient for SIZE was -11.52, but in the post-SOX period, the magnitude of coefficient for SIZE increased to -6.29 (= -11.52 + 5.23). The coefficient for SOX*GROWTH is significantly negative. The magnitude of the coefficient for GROWTH is reduced because SOX*GROWTH has a negative sign which lowers the positive coefficient on GROWTH in the post-SOX period. Thus, the magnitude of coefficient for GROWTH has been decreased from 6.18 to -4.58 (= 6.19-10.77) indicating that because of the increased monitoring, growth firms are more affected than value firms, which is consistent with hypothesis 4.

In order to test the direct impact of monitoring and increase the power of my tests, I use the firm as its own control in the next analysis and examine only the firms that issued debt in both pre-SOX and post-SOX periods. A total of 730 firms (2,940 facilities) issued debt in both periods. Table 6 shows that the main interest variable, SOX, is significantly negative, indicating that the cost of debt has decreased in the post-SOX period which is consistent with previous results and my hypotheses.

The Securities and Exchange Commission (SEC) allowed extended deadlines to July 15, 2007 for Sarbanes-Oxley compliance for small public companies (Wall Street Journal, September 22, 2005). The SEC defines a small company as one with a market capitalization of $700 million or less. Since the implementation of SOX was delayed for small public firms, I expect it will have less impact on these firms compared to the full sample. I find 1,694 observations that satisfy the condition of having market capitalization of $700 million or less. Table 7 shows that consistent with my expectation, the statistical evidence of SOX variable becomes weaker compared to Table 5 and Table 6 (i.e., t-statistic is only -2.37 in Table 7 compared to -4.94 and -3.03 in Table 5 and Table 6, respectively).

Instrumental Variable (IV) Approach

Debt contracting between two parties, lender and borrower, simultaneously determines cost of debt, debt covenants, maturity of debt, collateral, and amount of debt. Thus, the choice of one decision variable will affect the other, creating a potential endogeneity problem in the research design. It is well known that an endogeneity problem causes Ordinary Least Squares (OLS) regressions to be biased and inconsistent (Wooldridge, 2002). If the equations are estimated by OLS, there is potentially a problem of simultaneity bias because changes in maturity, collateral, amount, and debt covenant variables may affect one another and also affect the dependent variable. In addition, these explanatory variables could be correlated with the error term due to measurement error. This misspecification causes the parameter estimates to be inconsistent, which weakens the interpretation of the results presented in Table 5. Therefore, I use the instrumental variables (IV) approach to address the joint determination of the debt characteristic variables and to check the robustness of the basic results. Specifically, I employ the Two-Stage Least Squares (2SLS) method as follows: I first regress endogenous variables (debt characteristics) on each of the instruments and the exogenous variables (firm characteristics) to obtain fitted values for the endogenous variables (i.e., debt's maturity, borrowing amount, collateral, and debt covenant). Then, I substitute these fitted values for the endogenous variables and estimate the coefficient of the full model. To summarize, I obtain fitted values using a reduced form regression of debt characteristics in equation (4), then use equation (5) to obtain an unbiased and consistent estimator.

DC = f(FC,IV) (4)

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (5)

where DC represents debt characteristic variables, FC represents firm characteristic variables, IV represents an instrument, and [??[??]] represents fitted values of debt characteristic variables obtained from equation (4).

Loan agreements contain contractual provisions called debt covenants that require the borrower to maintain minimum levels of working capital, interest coverage, or other key accounting-based measures that provide a safety net to the lender. Debt covenants are clearly related to the credit characteristics of the borrower. Therefore, I use the current ratio as an instrumental variable for debt covenants. The most important requirement for using instrument variables is that z should be correlated with x. Instrumental variables are highly correlated with endogeneous variables. There correlations are 0.0539 (debt covenant and current ratio), 0.0736 (collateral and inverse of PPE), 0.2184 (maturity and yield curve), and 0.1772 (amount borrowed and R&D expense). I also tried other instruments but none of them satisfied the correlation requirement. For example, I used debt to cash flow, interest coverage, debt to equity for debt covenants, tangible asset ratio for collateral, and financing needs (following Jalilvand and Harris, 1984) for amount borrowed.

As stated by Dichev and Skinner (2002), the current ratio is the most standardized, unambiguous accounting measure and one of the most frequently violated debt covenants. For firms that have debt covenants in their debt contracts, violations of those debt covenants will result in renegotiation costs and increases in the cost of debt capital. Firms without debt covenants in their debt contracts will not necessarily suffer such repercussions if other factors are not in place. Therefore, the firm's current ratio and the existence of debt covenants will be highly correlated because the presence of debt covenants results in a higher probability that firms will suffer financially due to renegotiation costs and the increased cost of debt. Firms that have debt covenants, then, should be more careful not to violate these covenants by maintaining higher levels of current ratio. In addition, debt covenants generally require firms to meet certain financial ratios, such as a minimum current ratio level. Since current ratio covenants are the most frequently used covenants, the current ratio of a firm that has debt covenants in its debt contract is likely to be higher than the current ratio of a firm that does not have debt covenants in its debt contract.

Tangible assets like equipments, buildings, and lands can be used as collateral in debt contracts. Banks require collateral when the firm does not have enough tangible assets. Since debtholders have first claims to a firm's tangible assets, debtholders could liquidate the firm and get their portion of the tangible assets in case of bankruptcy. Thus, if the firm has enough tangible assets, debtholders have no need to worry about securing the loan. Therefore, I use the inverse of property, plant, and equipment as an instrumental variable for collateral.

Next, I use the yield curve as an instrument for debt maturity. The yield curve is normally upward sloping. That is, the interest rates for more distant maturities are normally higher than the interest rates for closer maturities, due to the risks associated with time. Thus, as maturity gets longer, the yield (i.e., Treasury bond rate) tends to be higher. Shortening loan maturity limits the risk, which results in lower spreads and fewer covenants.

Finally, I use research and development expense as an instrument for the amount of the loan. Firms that spend more on their research and development have more financing needs and thus borrow more. Firms generally engage in research and development to generate future investments.

I present instrumental variable approach estimation results in Table 8. The inferences from the 2SLS analyses are similar to those derived from the multiple regression reported in Table 5. The results show that the coefficient of SOX, the main variable of interest, is again significantly negative at the 1% level. (To address possible industry effects due to using R&D expense as an instrument, I repeat the test using two-digit SIC codes as an additional control variable on the first stage regression. However, the results remain unchanged.) Consistent with the multiple regression results, the cost of debt appears to have decreased significantly during the post-SOX period. Thus, the endogeneity problem does not affect the primary inferences drawn from Table 5, suggesting that the empirical results are robust to both procedures. However, the adjusted R-square has decreased from 34.73% to 15.15% in the Full Model (35.70% to 16.83% in Extended Model) compared to the OLS approach. One possible explanation for this is that it is difficult to find instruments that are perfectly correlated with endogeneous variables in the OLS regression.

CONCLUSION

This paper investigates the association among monitoring, debt covenants and the cost of debt for a sample of 4,610 private debt contracts that were issued between 1999 and 2005. While previous studies on debt contracts document only the association between monitoring and debt covenants (Black et al. 2004) or the association between debt covenants and the cost of debt (Beatty et al. 2002), I examine the association among all three (monitoring, debt covenants and cost of debt) and how they were affected by certain provisions of Sarbanes-Oxley Act of 2002.

I find evidence that, on average, during the post-SOX period where monitoring increased relative to the pre-SOX period, the cost of debt decreased. However, I do not find evidence that the number of debt covenants decreased during the post-SOX period. These results suggest that lenders substitute monitoring for borrowing costs. In addition, they suggest that the debtholders are willing to maintain debt covenants at a certain level.

I also examine the types of firms that are more likely to be influenced by increased monitoring during the post-SOX period and find that small firms and growth firms in particular are most affected, resulting in greater relative reductions in their cost of debt. These results are consistent with my hypothesis that firms with higher information asymmetry were significantly more affected by Sarbanes-Oxley. In addition, I provide evidence that small public companies with market capitalization of $700 million or less were only marginally affected by increased monitoring right after the passage of the Act, consistent with the SEC's decision to delay implementation for these firms.

Overall, the analysis in this paper provides evidence that the Sarbanes-Oxley Act of 2002 not only influenced the stock market but also the debt market. As such, it contributes to the literatures that examine the agency costs of debt and the economic impacts of the Sarbanes-Oxley Act and provides a foundation for future work on the relations between monitoring, debt covenants and the cost of debt.

REFERENCES

Barclay, M., Smith, C. (1995). The priority structure of corporate liabilities. Journal of Finance 50, 899-918.

Beatty, A., Weber, J. (2005). Performance pricing in debt contracts. Journal of Accounting and Economics 40, 101-128.

Beatty, A., Ramesh, K., Weber, J. (2002). The importance of accounting changes in debt contracts: The cost of flexibility in covenant calculations. Journal of Accounting and Economics 33, 205-227.

Begley, J., Feltham, G. A. (1999). An empirical examination of the relation between debt contracts and management incentives. Journal of Accounting and Economics 27, 229-259.

Berger, A., Udell, G. (1990). Collateral, loan quality and bank risk. Journal of Monetary Economics 25, 21-42.

Berger, P., Li, F., Wong, M. (2005). The impact of Sarbanes-Oxley on cross-listed companies. Working paper, University of Chicago.

Bharath, S., Dahiya, S., Saunders, A., Srinivasan, A. (2006). Lending relationships and loan contract terms: Does size matter? Working paper, New York University.

Black, E., Carnes, T., Mosebach, M., Moyer, S. (2004). Regulatory monitoring as a substitute for debt covenants. Journal of Accounting and Economics 37, 367-391.

DeFond, M. L., Jiambalvo, J. (1994). Debt covenant violation and manipulation of accruals. Journal of Accounting and Economics 17, 145-176.

Denis, D., Mihov, V. (2003). The choice among bank debt, non-bank private debt, and public debt: Evidence from new corporate borrowings. Journal of Financial Economics 70, 3-28.

Dichev, I., Skinner, D. (2002). Large-sample evidence on the debt covenant hypothesis. Journal of Accounting Research 40, 1091-1123.

Dixon, L., Gates, S. M., Kapur, K., Seabury, S. A., Talley, E. (2006). The impact of regulation and litigation on small business and entrepreneurship. Working paper, RAND Institute for Civil Justice.

Duke, J., Hunt, H. (1990). An empirical examination of debt covenant restrictions and accounting-related debt proxies. Journal of Accounting and Economics 12, 45-63.

Finnerty, J. (1976). Insiders and market efficiency. Journal ofFinance 31, 1141-1148.

Jain, P., Rezaee, Z. (2004). The Sarbanes-Oxley Act of 2002 and accounting conservatism. Working paper, University of Memphis.

Jalilvand, A., Harris, R. S. (1984). Corporate behavior in adjusting to capital structure and dividend targets: An econometric study. Journal of Finance 39, 127-145.

Jensen, M., Meckling, W. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics 3, 305-360.

John, K., Lynch, A. W., Puri, M. (2003). Credit rating, collateral and loan characteristics: Implications for yield. The Journal of Business 76, 371-409.

Krishnaswami, S., Spindt, P., Subramaniam, V. (1999). Information asymmetry, monitoring, and the placement structure of corporate debt. Journal of Financial Economics 51, 407-434.

Kwan, S., Carleton, W. T. (2004). Financial contracting and the choice between private placement and publicly offered bonds. Working paper, Federal Reserve Bank of San Francisco.

Lakonishok, J., Lee, I. (2001). Are insiders' trades informative? Review of Financial Studies 14, 79-111.

Li, H., Pincus, M., Rego, S. (2004). Market reaction to events surrounding the Sarbanes-Oxley Act of 2002. Working paper, University of Iowa.

Myers, S., C. (1977). Determinants of corporate borrowing. Journal of Financial Economics 5, 147-175.

Press, E., Weintrop, J. (1990). Accounting-based constraints in public and private debt agreements: Their association with leverage and impact on accounting choice. Journal of Accounting and Economics 12, 65-95.

Smith, C., Watts, R. (1992). The investment opportunity set and corporate financing, dividend, and compensation policies. Journal of Financial Economics 32, 263-292.

Seyhun, H. (1986). Insiders' profits, costs of trading, and market efficiency. Journal of Financial Economics 16, 189-212.

Zhang, I. X. (2007). Economic Consequences of the Sarbanes-Oxley Act of 2002. Journal of Accounting and Economics 44, 74-115.

Sangshin (Sam) Pae, Arkansas State University

Table 1. Sample selection
This table presents the sequential filters applied to obtain the
final sample of firm-quarters

 Number of
 Observations
 Remaining

Number of facilities during the sample period 35,358
Facilities with ticker symbols 13,962
Combine with quarterly Compustat 8,346
Combine with annual Compustat 4,668
Elimination of outliers in GROWTH variable 4,610

Table 2. Descriptive statistics for sample firm-quarters

Panel A. Descriptive statistics for regression variables

This table provides descriptive statistics for variables used in
subsequent tests. To be included in this table, a firm-quarter
observation must be accompanied by sufficient data to compute the
variables displayed below. The statistics for all variables are
based on 4,610 firm-quarter observations. Firm-quarter observations
are drawn from the period between 1999 and 2005, excluding 2002.

 Standard
 Obs. Mean Deviation

Spread 4,610 173.914 134.206
Size (in millions) 4,610 7335.5 14186.8
Leverage 4,610 0.2785 0.1612
Earnings Variability 4,610 93.4753 542.569
Market-to-Book Ratio 4,610 3.1285 5.4069
Facility Amount (in millions) 4,610 384.95 792.38
Maturity 4,610 3.4592 1.9843
Moody's Rating 4,610 6.2460 1.3483
Collateral 4,610 0.4440 0.4969
Debt Covenants 4,610 0.7536 0.4310
5-year T-bond Rate 4,610 4.3160 1.1458

 Lower Upper
 Quartile Median Quartile

Spread 62.5 150 250
Size (in millions) 500.411 2031.9 8248.05
Leverage 0.1654 0.2693 0.3780
Earnings Variability 1.9217 6.4552 28.1173
Market-to-Book Ratio 1.3076 2.0152 3.3740
Facility Amount (in millions) 55.00 175.00 425.00
Maturity 1.0833 3.1727 5
Moody's Rating 6 7 7
Collateral 0 0 1
Debt Covenants 1 1 1
5-year T-bond Rate 3.43 4.08 5.07

Table 2. Descriptive statistics for sample firm-quarters

Panel B. Difference in characteristics between pre-SOX and post-SOX

 Pre-SOX

 Standard
 Mean Median Deviation

Spread 169.597 140 131.888
Size (in millions) 7050.59 1873.63 13677.18
Leverage 0.2739 0.2657 0.1606
Earnings Variability 79.606 6.758 471.379
Market-to-Book Ratio 3.077 1.916 4.7149
Facility Amt (in millions) 365.25 150 935.84
Maturity 3.0489 2.9979 2.1189
Moody's Rating 6.2898 7 1.3898
Collateral 0.4288 0 0.4950
Debt Covenants 0.7186 1 0.4498
5-year T-bond Rate 5.3602 5.32 0.7810

 Post-SOX

 Standard
 Mean Median Deviation

Spread 177.349 150 135.949
Size (in millions) 7562.25 2150.5 14578.33
Leverage 0.2823 0.2728 0.1617
Earnings Variability 104.513 6.3012 593.012
Market-to-Book Ratio 3.170 2.0578 5.9005
Facility Amt (in millions) 400.63 200 655.93
Maturity 3.7857 4.1676 1.8052
Moody's Rating 6.2111 7 1.3136
Collateral 0.4562 0 0.4982
Debt Covenants 0.7815 1 0.4133
5-year T-bond Rate 3.4849 3.52 0.5601

 Test of Difference

 t-statistics p-value

Spread -1.96 0.0506
Size (in millions) -1.23 0.2205
Leverage -1.76 0.0792
Earnings Variability -1.59 0.1122
Market-to-Book Ratio -0.59 0.5525
Facility Amt (in millions) -1.45 0.1474
Maturity -12.51 <.0001
Moody's Rating 1.95 0.0507
Collateral -1.86 0.0630
Debt Covenants -4.89 <.0001
5-year T-bond Rate 91.42 <.0001

Table 3. Pearson correlation matrix

This table provides the value of the correlation between each of the
variables used in subsequent tests. To be included in this table, a
firm-quarter observation must be accompanied by sufficient data to
compute the variables displayed below. Therefore, the statistics for
all variables are based on 4,610 firm-quarter observations. Firm-
quarter observations are drawn from the period between 1999 and
2005, excluding 2002. ***, **, and * denote two-tailed significance
at the 0.01, 0.05, and 0.1 levels, respectively. Variable
definitions are as follows: SPREAD = basis point spread over LIBOR,
inclusive of all fees. SOX = indicator variable set equal to one if
the sample period is in the pre-SOX period; zero if the sample
period is in the post-SOX period. SIZE = natural log of total assets
of firm at quarter-end. LEV = ratio of long-term debt to total
assets at quarter-end. VAR = 5 years earnings variability prior to
debt contract; computed as the standard deviation of firm i's net
income before extraordinary items (scaled by total assets) measured
over rolling five year windows. GROWTH = market-to-book ratio;
computed as market value of equity divided by total book value of
equity. RATING = Moody's senior debt ratings information on each
firm; ratings of Aaa, Aa, A, Baa, Ba, B, and lower than B represents
1 through 7, respectively. MAT = stated maturity computed in years.
AMOUNT = amount of money borrowed in debt contract scaled by total
assets. COL = dummy variable set equal to one if the loan is
secured, zero otherwise. COV = dummy variable set equal to one if
the debt covenant is included in debt contract, zero otherwise.
TBOND = 5 year Treasury bond rate.

 SPREAD SOX SIZE LEV

SPREAD 1.000
SOX 0.0290 1.000
 0.0514
SIZE -0.265 *** 0.018 1.000
 <.0001 0.2239
LEV 0154 *** 0.0260 0.098 *** 1.000
 <.0001 0.0794 <.0001
VAR -0.040 *** 0.023 0.014 -0.055 ***
 0.0064 0.1216 0.3418 0.0002
GROWTH 0.098 *** -0.016 -0.008 0.012
 <.0001 0.2834 0.5824 0.4085
RATING 0.123 *** -0.029 ** 0.096 *** 0.011
 <.0001 0.0492 <.0001 0.4484
MAT 0.173 *** 0.184 *** -0.211 *** 0.145 ***
 <.0001 <.0001 <.0001 <.0001
AMOUNT -0.261 *** 0.002 0.240 *** 0.002
 <.0001 0.132 <.0001 0.8955
COL 0.548 *** 0.0270 -0.292 *** 0.104 ***
 <.0001 0.063 <.0001 <.0001
COV 0.205 *** 0.073 *** -0.134 *** 0.044 ***
 <.0001 <.0001 <.0001 0.0029
TBOND -0.053 *** -0.813 *** -0.167 *** -0.0270
 0.0003 <.0001 <.0001 0.0622

 VAR GROWTH RATING MAT

SPREAD
SOX

SIZE

LEV

VAR 1.000

GROWTH -0.014 1.000
 0.3304
RATING -0.054 *** 0.045 *** 1.000
 0.0003 0.0021
MAT -0.021 0.011 0.005 1.000
 0.1611 0.4484 0.7386
AMOUNT 0.024 -0.0290 -0.101 *** -0.074 ***
 0.1023 0.0504 <.0001 <.0001
COL -0.0250 0.005 0.111 *** 0.261 ***
 0.0853 0.7369 <.0001 <.0001
COV -0.009 -0.035 ** 0.059 *** 0.165 ***
 0.5327 0.0164 <.0001 <.0001
TBOND -0.023 0.004 -0.073 *** -0.072 ***
 0.1176 0.8071 <.0001 <.0001

 AMOUNT COL COV TBOND

SPREAD
SOX

SIZE

LEV

VAR

GROWTH

RATING

MAT

AMOUNT 1.000

COL -0.217 *** 1.000
 <.0001
COV -0.066 *** 0.427 *** 1.000
 <.0001 <.0001
TBOND -0.004 -0.030 -0.085 *** 1.000
 0.7756 0.0446 <.0001

Table 4. Logit regression results based on 4,610 firm-quarter
observations

This table provides the results of logit regression with the
dependent variable COV. To be included in this table, a firm-
quarter observation must be accompanied by sufficient data to
compute the variables displayed below. Therefore, the statistics for
all variables are based on 4,610 firm-quarter observations. Firm-
quarter observations are drawn from the period between 1999 and
2005, excluding 2002. ***, **, and * denote two-tailed significance
at the 0.01, 0.05, and 0.1 levels, respectively. Variable
definitions are as follows: COV = dummy variable set equal to one if
the debt covenant is included in debt contract, zero otherwise. SOX
= indicator variable set equal to one if the sample period is in the
pre-SOX period; zero if the sample period is in the post-SOX period.
SIZE = natural log of total assets of firm at quarter-end. LEV =
ratio of long-term debt to total assets at quarter-end. VAR = 5
years earnings variability prior to debt contract; computed as the
standard deviation of firm i's net income before extraordinary items
(scaled by total assets) measured over rolling five year windows.
GROWTH = market-to-book ratio; computed as market value of equity
divided by total book value of equity. RATING = Moody's senior debt
ratings information on each firm; ratings of Aaa, Aa, A, Baa, Ba, B,
and lower than B represents 1 through 7, respectively. MAT = stated
maturity computed in years. AMOUNT = amount of money borrowed in
debt contract scaled by total assets. COL = dummy variable set equal
to one if the loan is secured, zero otherwise. SPREAD = basis point
spread over LIBOR, inclusive of all fees. TBOND = 5 year Treasury
bond rate.

[Prob[COV=1] = logit([[alpha].sub.0] + [[alpha].sub.1][SOX.sub.i,t]
+ [[alpha].sub.2][SIZE.sub.i,t] + [[alpha].sub.3][LEV.sub.i,t] +
[[alpha].sub.4][VAR.sub.i,t] + [[alpha].sub.5][GROWTH.sub.i,t] +
[[alpha].sub.6][RATING.sub.i,t] + [[alpha].sub.7][MAT.sub.i,t] +
[[alpha].sub.8][AMOUNT.sub.i,t] + [[alpha].sub.9][COL.sub.i,t] +
[[alpha].sub.10][SPREAD.sub.i,t] + [[alpha].sub.11][TBOND.sub.i,t]

 Dependent Variable

Independent Predicted COV
Variables Sign Coefficient Chi-Square

Intercept 1.299 9.40 ***
SOX - -0.122 0.78
SIZE - -0.025 2.33
LEV +/- 0.093 0.12
VAR 0 0.000 0.00
GROWTH 0 -0.035 1.67
RATING 0 0.032 1.39
MAT 0 0.075 12.96 ***
AMOUNT +/- -0.002 0.41
COL - 2.896 462.15 ***
SPREAD - -0.001 12.45 ***
TBOND ? -0.224 14.26 ***
Obs. 4,610
Chi-square 5.1910
p-value 0.7370
R-square 0.2030

Table 5. Multiple regression results based on 4,610 firm-quarter
observations

This table provides the results of multiple regression with the
dependent variable SPREAD. To be included in this table, a firm-
quarter observation must be accompanied by sufficient data to
compute the variables displayed below. Therefore, the statistics for
all variables are based on 4,610 firm-quarter observations. Firm-
quarter observations are drawn from the period between 1999 and
2005, excluding 2002. **, and * denote two-tailed significance at
the 0.01, 0.05, and 0.1 levels, respectively. Variable definitions
are as follows: SPREAD=basis point spread over LIBOR, inclusive of
all fees. SOX = indicator variable set equal to one if the sample
period is in the pre-SOX period; zero if the sample period is in the
post-SOX period. SIZE=natural log of total assets of firm at
quarter-end. LEV = ratio of long-term debt to total assets at
quarter-end. VAR = 5 years earnings variability prior to debt
contract; computed as the standard deviation of firm i's net income
before extraordinary items (scaled by total assets) measured over
rolling five year windows. GROWTH = market-to-book ratio; computed
as market value of equity divided by total book value of equity.
RATING = Moody's senior debt ratings information on each firm;
ratings of Aaa, Aa, A, Baa, Ba, B, and lower than B represents 1
through 7, respectively. MAT=stated maturity computed in years.
AMOUNT = amount of money borrowed in debt contract scaled by total
assets. COL=dummy variable set equal to one if the loan is secured,
zero otherwise. COV = dummy variable set equal to one if the debt
covenant is included in debt contract, zero otherwise. TBOND = 5
year Treasury bond rate.

[SPREAD.sub.i,t] = [[beta].sub.0] + [[beta].sub.1][SOX.sub.i,t] +
[[beta].sub.2][SIZE.sub.i,t] + [[beta].sub.3][LEV.sub.i,t] +
[[beta].sub.4][VAR.sub.i,t] + [[beta].sub.5][GROWTH.sub.i,t] +
[[beta].sub.6][RATING.sub.i,t] + [[beta].sub.7][MAT.sub.i,t] +
[[beta].sub.8][AMOUNT.sub.i,t] + [[beta].sub.9][COL.sub.i,t] +
[[beta].sub.10][COV.sub.i,t] + [[beta].sub.11][TBOND.sub.i,t]

 Independent Predicted Model (1)
 Variables Sign

 Coefficient t-stat

Intercept 169.597 57.14 ***
SOX - 7.752 1.950
SIZE -
LEV +/-
VAR 0
GROWTH 0
RATING 0
MAT 0
AMOUNT +/-
COL -
COV -
TBOND ?
SOX*SIZE 0
SOX*GROWTH -
Obs. 4,610
Adj. R-square 0.0006

 Independent Full Model
 Variables

 Coefficient t-stat

Intercept 199.696 10.69 ***
SOX -29.160 -4.94 ***
SIZE -9.014 -12.83 ***
LEV 97.078 9.54 ***
VAR -0.004 -1.29
GROWTH 3.757 7.28 ***
RATING 6.830 5.61 ***
MAT 0.360 0.41
AMOUNT -0.567 -2.99 ***
COL 132.310 34.73 ***
COV -12.439 -3.02 ***
TBOND -17.590 -6.83 ***
SOX*SIZE
SOX*GROWTH
Obs. 4,610
Adj. R-square 0.3473

 Independent Extended Model
 Variables

 Coefficient t-stat

Intercept 205.887 10.97 ***
SOX -50.588 -5.42 ***
SIZE -11.516 -12.05 ***
LEV 95.281 9.43 ***
VAR -0.004 -1.29
GROWTH 13.038 9.36 ***
RATING 6.187 5.11 ***
MAT 0.339 0.39
AMOUNT -0.596 -3.16 ***
COL 130.846 34.51 ***
COV -13.032 -3.18 ***
TBOND -16.111 -6.24 ***
SOX*SIZE 5.225 4.16 ***
SOX*GROWTH -10.772 -7.20 ***
Obs. 4,610
Adj. R-square 0.3570

Table 6. Multiple regression results based on 2,940 firm-quarter
observations (sample includes only the firms that issued both
pre-SOX and post-SOX period)

This table provides the results of multiple regression with the
dependent variable SPREAD. To be included in this table, a firm-
quarter observation must be accompanied by sufficient data to
compute the variables displayed below. Therefore, the statistics for
all variables are based on 2,940 firm-quarter observations. Firm-
quarter observations are drawn from the period between 1999 and
2005, excluding 2002. ***, **, and * denote two-tailed significance
at the 0.01, 0.05, and 0.1 levels, respectively. Variable
definitions are as follows: SPREAD=basis point spread over LIBOR,
inclusive of all fees. SOX = indicator variable set equal to one if
the sample period is in the pre-SOX period; zero if the sample
period is in the post-SOX period. SIZE = natural log of total assets
of firm at quarter-end. LEV=ratio of long-term debt to total assets
at quarter-end. VAR = 5 years earnings variability prior to debt
contract; computed as the standard deviation of firm i's net income
before extraordinary items (scaled by total assets) measured over
rolling five year windows. GROWTH = market-to-book ratio; computed
as market value of equity divided by total book value of equity.
RATING = Moody's senior debt ratings information on each firm;
ratings of Aaa, Aa, A, Baa, Ba, B, and lower than B represents 1
through 7, respectively. MAT = stated maturity computed in years.
AMOUNT = amount of money borrowed in debt contract scaled by total
assets. COL = dummy variable set equal to one if the loan is
secured, zero otherwise. COV = dummy variable set equal to one if
the debt covenant is included in debt contract, zero otherwise.
TBOND = 5 year Treasury bond rate.

[SPREAD.sub.i,t] = [[gamma].sub.0] + [[gamma].sub.1][SOX.sub.i,t] +
[[gamma].sub.2][SIZE.sub.i,t] + [[gamma].sub.3][LEV.sub.i,t] +
[[gamma].sub.4][VAR.sub.i,t] + [[gamma].sub.5][GROWTH.sub.i,t] +
[[gamma].sub.6][RATING.sub.i,t] + [[gamma].sub.7][MAT.sub.i,t] +
[[gamma].sub.8][AMOUNT.sub.i,t] + [[gamma].sub.9][COL.sub.i,t] +
[[gamma].sub.10][COV.sub.i,t] + [[gamma].sub.11][TBOND.sub.i,t] +
[[gamma].sub.12][(SOX*SIZE).sub.i,t] + [[gamma].sub.13]
[(SOX*GROWTH).sub.i,t]

Independent Predicted Full Model
Variables Sign
 Coefficient t-stat

Intercept 180.734 8.48 ***
SOX - -20.658 -3.03 ***
SIZE - -10.588 -11.64 ***
LEV +/- 122.401 9.29 ***
VAR 0 -0.002 -0.64
GROWTH 0 2.577 4.91 ***
RATING 0 6.489 4.80 ***
MAT 0 0.782 0.75
AMOUNT +/- -13.334 -5.98 ***
COL - 131.255 27.86 ***
COV - -9.331 -2.00 **
TBOND ? -13.964 -4.74 ***
SOX*SIZE 0
SOX*GROWTH -
Obs. 2,940
Adj. R-square 0.3780

Independent Extended Model
Variables
 Coefficient t-stat

Intercept 178.298 8.21 ***
SOX -35.383 -2.91 ***
SIZE -12.747 -10.71 ***
LEV 124.838 9.54 ***
VAR -0.002 -0.62
GROWTH 25.245 6.60 ***
RATING 5.553 4.12 ***
MAT 1.222 1.18
AMOUNT -12.939 -5.83 ***
COL 127.763 27.16 ***
COV -10.908 -2.36 **
TBOND -12.378 -4.21 ***
SOX*SIZE 4.848 3.17 ***
SOX*GROWTH -23.110 -5.99 ***
Obs. 2,940
Adj. R-square 0.3874

Table 7. Multiple regression results based on 1,694 firm-quarter
observations (sample includes only the small firms with a market
capitalization of $700 million or less)

This table provides the results of multiple regression with the
dependent variable SPREAD. To be included in this table, a firm-
quarter observation must be accompanied by sufficient data to
compute the variables displayed below. Therefore, the statistics for
all variables are based on 1,694 firm-quarter observations. Firm-
quarter observations are drawn from the period between 1999 and
2005, excluding 2002. ***, **, and * denote two-tailed significance
at the 0.01, 0.05, and 0.1 levels, respectively. Variable
definitions are as follows: SPREAD=basis point spread over LIBOR,
inclusive of all fees. SOX = indicator variable set equal to one if
the sample period is in the pre-SOX period; zero if the sample
period is in the post-SOX period. SIZE = natural log of total assets
of firm at quarter-end. LEV = ratio of long-term debt to total
assets at quarter-end. VAR = 5 years earnings variability prior to
debt contract; computed as the standard deviation of firm i's net
income before extraordinary items (scaled by total assets) measured
over rolling five year windows. GROWTH = market-to-book ratio;
computed as market value of equity divided by total book value of
equity. RATING = Moody's senior debt ratings information on each
firm; ratings of Aaa, Aa, A, Baa, Ba, B, and lower than B represents
1 through 7, respectively. MAT = stated maturity computed in years.
AMOUNT = amount of money borrowed in debt contract scaled by total
assets. COL = dummy variable set equal to one if the loan is
secured, zero otherwise. COV = dummy variable set equal to one if
the debt covenant is included in debt contract, zero otherwise.
TBOND = 5 year Treasury bond rate.

[SPREAD.sub.i,t] = [[gamma].sub.0] + [[gamma].sub.1][SOX.sub.i,t] +
[[gamma].sub.2][SIZE.subl.i,t] + [[gamma].sub.3][LEV.sub.i,t] +
[[gamma].sub.4][VAR.sub.i,t] + [[gamma].sub.5][GROWTH.sub.i,t] +
[[gamma].sub.6][RATING.sub.i,t] + [[gamma].sub.7][MAT.sub.i,t] +
[[gamma].sub.8][AMOUNT.sub.i,t] + [[gamma].sub.9][COL.sub.i,t] +
[[gamma].sub.10][COV.sub.i,t] + [[gamma].sub.11][TBOND.sub.i,t] +
[[gamma].sub.12][(SOX*SZE).sub.i,t] +
[[gamma].sub.13][(SOX*GROWTH).sub.i,t]

 Independent Predicted Full Model
 Variables Sign
 Coefficient t-stat

Intercept 117.849 2.48 **
SOX - -25.197 -2.37 **
SIZE - 3.843 2.41 **
LEV +/- 59.684 3.69 ***
VAR 0 -0.008 -0.70
GROWTH 0 2.583 4.63 ***
RATING 0 17.132 3.43 ***
MAT 0 -1.310 -0.84
AMOUNT +/- -0.499 -2.46 **
COL - 126.934 18.13 ***
COV - -32.665 -3.68 ***
TBOND ? -16.956 -3.71 ***
SOX*SIZE 0
SOX*GROWTH
Obs. 1,694
Adj. R-square 0.2304

 Independent Extended Model
 Variables
 Coefficient t-stat

Intercept 127.585 2.73 ***
SOX -78.188 -4.80 ***
SIZE -3.598 -1.75 0
LEV 57.123 3.59 ***
VAR -0.010 -0.81
GROWTH 11.541 7.49 ***
RATING 15.977 3.25 ***
MAT -1.455 -0.95
AMOUNT -0.541 -2.71 ***
COL 123.344 17.82 ***
COV -29.474 -3.37 ***
TBOND -12.999 -2.84 ***
SOX*SIZE 15.646 5.19 ***
SOX*GROWTH -10.313 -6.27 ***
Obs. 1,694
Adj. R-square 0.2564

Table 8. Instrumental Variable approach results based on 4,610 firm-
quarter observations

This table provides the results of instrumental variable estimation
approach with the dependent variable SPREAD. To be included in this
table, a firm-quarter observation must be accompanied by sufficient
data to compute the variables displayed below. Therefore, the
statistics for all variables are based on 4,610 firm-quarter
observations. Firm-quarter observations are drawn from the period
between 1999 and 2005, excluding 2002. ***, **, and * denote two-
tailed significance at the 0.01, 0.05, and 0.1 levels, respectively.
Variable definitions are as follows: SPREAD = basis point spread
over LIBOR, inclusive of all fees. SOX = indicator variable set
equal to one if the sample period is in the pre-SOX period; zero if
the sample period is in the post-SOX period. SIZE = natural log of
total assets of firm at quarter-end. LEV = ratio of long-term debt
to total assets at quarter-end. VAR = 5 years earnings variability
prior to debt contract; computed as the standard deviation of firm
i's net income before extraordinary items (scaled by total assets)
measured over rolling five year windows. GROWTH = market-to-book
ratio; computed as market value of equity divided by total book
value of equity. RATING = Moody's senior debt ratings information on
each firm; ratings of Aaa, Aa, A, Baa, Ba, B, and lower than B
represents 1 through 7, respectively. fit(MAT) = fitted value of
maturity using yield curve as an instrument. fit(AMOUNT) = fitted
value of amount borrowed using research and development as an
instrument. fit(COL) = fitted value of collateral using inverse of
property, plant, and equipment as an instrument. fit(COV) = fitted
value of debt covenant using current ratio as an instrument. TBOND = 5
year Treasury bond rate.

[SPREAD.sub.i,t] = [[upsilon].sub.0] +
[[upsilon].sub.1][SOX.sub.i,t] + [[upsilon].sub.2][SIZE.sub.i,t] +
[[upsilon].sub.3][LEV.sub.i,t] + [[upsilon].sub.0][VAR.sub.i,t] +
[[upsilon].sub.5][GROWTH.sub.i,t] +
[[upsilon].sub.6][RATING.sub.i,t] +
[[upsilon].sub.7]fit[(MAT).sub.i,t] +
[[upsilon].sub.8]fit[(AMOUNT).sub.i,t] +
[[upsilon].sub.9]fit[(COL).sub.i,t] +
[[upsilon].sub.10]fit[(COV).sub.i,t] +
[[upsilon].sub.11][TBOND.sub.i,t] +
[[upsilon].sub.12][(ZOX*SIZE).sub.i,t] +
[[upsilon].sub.13][(SOX*GROWTH).sub.i,t]

 Independent Predicted Full Model
 Variables Sign
 Coefficient t-stat

Intercept 69.085 14.36 ***
SOX - -47.473 -7.03 ***
SIZE - -16.087 -19.53 ***
LEV +/- 148.006 12.45 ***
VAR 0 -0.005 -1.38
GROWTH 0 3.645 6.20 ***
RATING 0 11.850 8.25 ***
fit(MAT) 0 48.584 4.37 ***
fit(AMOUNT) +/- -0.008 -0.57
fit(COL) - 39.072 0.74
fit(COV) - 164.390 1.95 0
TBOND ? -44.604 -8.65 ***
SOX*SIZE 0
SOX*GROWTH -

Obs. 4,610
Adj. R-square 0.1515

 Independent Extended Model
 Variables
 Coefficient t-stat

Intercept 101.061 15.12 ***
SOX -96.935 -9.05 ***
SIZE -20.972 -19.24 ***
LEV 143.554 12.19 ***
VAR -0.005 -1.36
GROWTH 13.132 8.29 ***
RATING 11.077 7.77 ***
fit(MAT) 54.717 4.96 ***
fit(AMOUNT) -0.003 -0.23
fit(COL) 13.837 0.27
fit(COV) 138.085 1.65 0
TBOND -43.982 -8.61 ***
SOX*SIZE 10.070 7.04 ***
SOX*GROWTH -11.044 -6.49 ***

Obs. 4,610
Adj. R-square 0.1683