“Debt maturity and corporate R&D investment – the empirical study of US listed firms”

This study investigates the relationships between debt maturity structure and corporation R&D investment. Using a large sample of US listed firms over the period of 1995 to 2015, it was found that the use of bank debt positively influences R&D investment, whereas the use of public debt exerts a negative impact. However, the Sarbanes-Oxley Act (SOX) mitigates the information asymmetry such that the advantages of private information from banks shrunk. As a result, public debtholders benefit more from the SOX and turn out to be positively influenced by the R&D investment after SOX. Moreover, bank debt impact on R&D spending reduces over the post-SOX. The results also find that the SOX influences the debt maturity on corporate R&D investment only for large corporations, the effects remain unchanged for small businesses.


INTRODUCTION
According to the Loan Pricing Corporation and the Federal Reserve System, the total amount of equity issued was approximately $115 billion, the total amount of corporate bonds issued was roughly $700 billion, and the total amount of bank loans issued was nearly $1,500 billion in 2005 (Chi, Huang, & Xie, 2015).
Knowledge, technical progress, and human capital exert a substantial impact on firm growth and survival. Consequently, uncovering the determinants of innovation is crucial for managers to foster R&D investment. Corporate R&D decision is part of the investment decisions for allocating the limited financial resources of firms toward different potential projects. External financing has always been a primary source of firm financing, and deciding between borrowing from banks or issuing bonds or equity is challenging. However, highly innovative companies have difficulty in acquiring external funding due to the asymmetric information problem between firms and outside investors (Kamien & Schwartz, 1978). This asymmetric information favors the issue of private debt, as businesses must share their private information with debtholders to reveal that the investment is profitable, to access to finance for innovative projects.
Since most previous studies on financing corporate innovation focus on internal financing or equity channels, very few papers mentioned debt financing. Among corporations with external financing, the amount of debt has exceeded the equity 1 . Given the economic importance of debt in allocating capital to firms, an emerging stream of research explores the relationship between leverage and investment decisions. For example, McConnell and Servaes (1995), Bhagat and Welch (1995), Aivazian et al. (2005) and Firth et al. (2008) examine the effect of leverage on firm investment and find that leverage significantly negatively influences corporate investment. However, research exploring the effect of debt maturity structure on R&D investment is rare. Moreover, whether shortterm bank borrowings and long-term public debt influence the pace of firm R&D investment remains an unresolved empirical issue.
The motivation of this work involves the facts that R&D spending is a high-risk investment and has the difficulty in financing such investment from outsiders due to asymmetric information (Kamien & Schwartz, 1978;Nofsinger & Wang, 2011). Innovative firms have lower incentive to release information regarding know-how, confidentiality, and competitiveness. Therefore, companies with high-quality projects may avoid issuing public debt given the high costs of information disclosure and potential loss of competitive advantage to competitors (Yosha, 1995). By contrast, bank debt is more advantageous than public debt in terms of keeping information private. Banks used to maintain long-term relationships with borrowing firms and accumulate soft information about the businesses (Diamond, 1991;Chemmanur & Fulghieri, 1994;Lin et al., 2013). Banks can also access inside information on borrowing firms, whereas outside debt holders must rely mostly on publicly available information (Fama, 1985;Boot, 2000). Although public creditors and banks have relative advantages for innovative companies, the tradeoff between them remains a concern for managers. How can an effective monitoring or less information exposure help foster the success in innovation? How does debt maturity influence a firm's innovation? All these issues deserve further examination. Aghion et al. (2004) show that firms having R&D innovation input tend to use more debt than those without R&D activities. They also find a non-linear relationship between innovation and debt finance.
Recently, some researchers examine the relationship between bank development/deregulation and corporate innovation outputs ( Given the limited empirical research on exploring the relationships between debt maturity structure and R&D investment in US firms, this work employs a measure for debt maturity. Short-term bank borrowings and long-term public debt to measure debt are used maturity structure. Biases due to firm heterogeneity and endogeneity present a concern in the analysis. The authors carefully choose the set of control variables and employ two-stage least squares (2SLS) regression to address the endogeneity issue. Further, multivariate difference-in-differences analysis is conducted to provide robust tests on the relationships among bank debt, public debt, and R&D investment. The results are quite consistent. The study contributes to the literature on capital structure and corporate innovation, especially on the angle of financing innovation through debt channels.
The findings reveal a positive relationship between the use of bank debt and R&D investment: a higher fraction of bank financing is associated with a greater level of R&D investment. Conversely, public debt use seems to mitigate R&D investment due to a negative relationship between public debt and R&D investment. Moreover, the enactment of the Sarbanes-Oxley Act (SOX) in 2002 mitigates the information asymmetry, such that public debtholders gain more benefits from the SOX than banks in the relationship with R&D investment. This paper presents implications for innovative firms making financing decisions.
The remainder of this paper is organized as follows. Section 2 briefly reviews the relevant literature. Section 3 describes the data and methodology, followed by the empirical results in Section 4 and the conclusions in the last Section.

LITERATURE REVIEW
Innovation involves the development of novel technological know-how to generate higher quality or lower cost product relative to extant resources (O'Sullivan, 2000). A number of notable scholars are in agreement regarding the value of innovation along with that of R&D investment as the keys to economic development and company growth (Zahra & Covin, 1995;Cohen & Klepper, 1996;Hurley & Hult, 1998;Woodside, 2005). R&D activities have high failure probability and uncertain return and thus are regarded as risky investment; nevertheless, they considerably benefit an enterprise (Bergemann & Hege, 2005).
Agency theory indicates that managers are risk averse due to concerns about their undiversified human capital (Fama, 1980). In response to this, later researches explore the means to motivate managers into making risky decisions by using external and internal corporate governance mechanisms. External mechanisms include supervision by the board of directors, shareholders, and lenders.
Diamond (1984) develops a theory of financial intermediation based on minimizing the cost of monitoring information to resolve incentive problems that arise between borrowers and lenders. It is expected that the lenders, particularly banks, can retrieve proprietary information and ensure effectiveness in terms of monitoring because of their roles as insider and delegated monitors. Modigliani and Miller (1958) are one of the first who investigated the role of debt. They later claim that the firm value increases with debt due to the two advantages of debt. One benefit is from tax shield, that is, interest payments are typically not taxed (Modigliani & Miller, 1963;Atiyet, 2012). The other advantage is that managers can be disciplined through debt. The law usually guarantees a right of information disclosure to debt holders, which serves as another tool for monitoring managers (Jensen, 1986). Consequently, debt disciplines managers, forcing them to disinvest, or to finance new investment, even in firms that have ready access to additional equity (Lambrecht & Myers, 2008) and create maximum value for the firms (Wruck, 1994). Jensen (1986) describes two agency problems resulting from debt financing, namely, underinvestment and overinvestment. He suggests that firms with substantial free cash flow have a tendency to invest more by accepting projects with low expected returns. In these cases, debts can be used to mitigate the effects of overinvestment because managers would use the excess cash flow to pay out the interest rather than invest in projects with negative net present value. Similarly, Gomariz and Ballesta (2014) examine the effect of debt maturity on investment efficiency for Spanish firms and indicate that short-term debt would improve investment efficiency, reduce both underinvestment and overinvestment problems. A recent study uses the dispersion of debt maturities and constitutes a crucial dimension to capital structure choice (Choi et al., 2018). It also inspires the authors to investigate the relationships between debt maturity structure and corporation R&D investment.
The question of how the use of debt should matter in innovation investments has been examined in prior literature, but results are mixed across various countries. Using US and Japanese firms as sample, Bhagat and Welch (1995) explore the determining factors of corporate R&D and find a significantly negative correlation between the debt ratio of the previous year and the present R&D expenses in US companies, although this ratio is significantly positively correlated for Japanese firms. Singh and Faircloth (2005) confirm a significantly negative relationship between the financial leverage and R&D expenditure. Their findings further indicate that high leverage leads to low R&D expenditure instead of vice versa. Empirical study by Czarnitzki and Kraft (2009) verifies that debt financing itself exerts a negative impact on firms' innovativeness in Germany. Conversely, employing data from manufacturing firms in Italy, Alfonso and Giannangeli (2012) find a positive effect of debt financing on the probability of R&D outsourcing. They argue that the asymmetric information problems plaguing the relationship between a firm and its lenders are particularly important in the financing of R&D.
The mixed regarding the effects of leverage and the gap in examining the influence of debt maturity structure on R&D investment leave room for us to add to the existing literature. In this research, the focus is on debt financing and investigating whether short-term bank debt or long-term public debt give more incentives for R&D investment. Critics suggest an important role of bank debt through lending, mitigating asymmetric information, and monitoring the financing of firm investments. Fama (1985) and Diamond (1991) indicate that the asymmetric information problem may be reduced because banks specialize in collecting the borrowing firm's private information and maintaining long-term relationships with those firms. Calomiris and Kahn (1991) address that short-term debt serves as a mechanism to supervise the managers. Jiménez and Saurina (2004) indicate that a close bank-borrower relationship increases the willingness of firm to take risk. Given these arguments, the authors of the article hypothesize a positive relationship between bank debt and firm R&D investment. Furthermore, the view of Yosha (1995) is followed who argues that firms with high-quality projects are likely to avoid public debt due to the high information disclosure costs and the potential loss of competitive advantage relative to rivals. It is expected that public debt is negatively related to R&D investment considering that R&D projects entail high-growth options for firms.

METHODS
The Compustat Fundamentals Annual database is used to collect the financial statements of US companies during 1995-2015. The R&D investment (RD_TA) was measured by R&D expenses divided by total assets. Leverage, debt maturity, firm size, age, cash ratio, return on assets, and growth opportunities are a set of control variables. The related definitions are presented in the next section and in

Measuring the variables
This section provides the definitions of the dependent and independent variables. The detailed variable definitions are also presented in Table 1a of the Appendix.
Research and development investments (RD_TA): A firm's R&D expenditure ratio during the fiscal year was measured as innovative investment.
RD_TA is defined as the ratio of R&D expenditure (Compustat item XRD) to the book value of the total assets (Compustat item AT) 2 .
Bank debt (BANKDEBT_TA): Bank debt is measured as the total amount of short-term borrowings, including bank advances, bank liabilities, bank overdrafts, current bank loans, notes payable to banks, and other borrowed funds reported as part of short-term borrowings. Bank debt ratio equals bank debt divided by the book value of the total assets (Compustat item AT). Given that banks help mitigate the asymmetric information problem, a positive relationship is expected between bank debt and firm R&D investment.
Public debt (PUBLICDEBT_TA): Long-term public debt is measured by the amount of total long-term debt, including long-term borrowings and corporate bonds. Public debt is calculated as the ratio of long-term debt (Compustat item DLTT) divided by the book value of the total assets (Compustat item AT). A negative relationship is expected between public debt and R&D investment.
Firm size (LNSALE): A firm's size and age affect innovation; thus, the size reflects a company's present situation and innovation prospects. Firm size is measured using the natural logarithm of total sale (Compustat item SALE) (Craig & Dibrell, 2006). Firm innovation is expected to be positively correlated with firm size.
Firm age (LNAGE): Firm age was measured by the natural logarithm of one plus the number of years as a publicly listed company. As the start-up phase is usually linked to the commercialization of inventions, younger firms may be viewed as more innovative than established firms. As described above, established firms might be more reluctant to introduce their innovations.
Return on assets (ROA): Profitability is reflected in this variable. Return on assets is defined as the net operating income (Compustat item NI) divided by the book value of the total assets (Compustat item AT). Hitt  Cash ratio (CASHRATIO): Cash ratio is measured by the ratio of cash and short-term market securities to total assets (Compustat item AT). The capital expenditure ratio, or CAPEX_TA, is measured by the capital expenditure (Compustat item CAPX) to total assets. Cash ratio and capital expenditure is expected to be positively related to R&D investment.
Industry effect: Two-digit SIC dummies are used to control for potential differences in firm innovations because opportunities differ among industries.
Time/Event effects: The variable SOX2002 controls for the impact of the SOX in 2002 to examine the shock on influencing the impact between debt maturity structure and firm R&D investment, which equals one if the year was after 2002, and zero otherwise. Extant literature verifies that highly innovative companies face difficulties in attracting capital for implementing projects (Bushee, 1998;Graves & Waddock, 1990). Moreover, although a high level of information asymmetry between firms and financiers may cause high costs of monitoring from outside financiers (Abdioglu et al., 2013;Jensen & Meckling, 1976), the SOX was believed to improve financial disclosure and the quality of information (Engel et al., 2007). As such, a positive relationship expected between public debt and R&D investment over the post-SOX period.

Methodology
In examining the relationship between the debt maturity structure and firm innovation investment, firm innovative investment (RD_TA) is first regressed on the debt structure that includes bank debt and public debt. The authors also control for firm operational characteristics and other industries fixed effects, and time and event factors. The regression is specified as equation (1) ed by the total assets. An estimated model is first built for BANKDEBT_TA and PUBLICDEBT_TA using the above instrumental variables and all the control variables in equation (1). The models are shown as equations (2) and (3) where the subscripts i and t indicate firm and time, respectively. ϒ is the error term. The error terms in both equations may be correlated. Therefore, adopting a 2SLS regression to address the endogeneity issue and the correlated errors between the equations is reasonable. The Hausman specification test (Hausman, 1978) is employed to confirm the existence of endogeneity. Further, the identification test and excluded-instruments F-test are also conducted to check for soundness and adequate instruments.

Descriptive statistics
Panels A and B of Table 1 report the descriptive statistics of the endogenous and exogenous variables employed in this study. Interestingly, on average, 46% of firms' financing is provided by bank loans and approximately 20% is from public debt. These results confirm the role of debt financing in the US firms as shown in existing literature. The  Note: Panels A, B, and C present the descriptive statistics and Panel D provides the Pearson correlation coefficient for the sample of US public firms from 1995 to 2015. RD_TA is the ratio of R&D expenditure to the book value of assets. BANKDEBT_TA is the ratio of short-term borrowing to total assets. PUBLICDEBT_TA is the ratio of long-term borrowing and corporate bonds to total assets. Firm size (LNSALE) is the natural logarithm of the total sales. Firm age (LNAGE) is the natural logarithm of one plus the number of years as a publicly listed company. CASHRATIO is the ratio of the total cash and cash equivalents to total assets. CAPEX_TA is the capital expenditure ratio, which is the capital expenditure divided by the total assets. Profitability (ROA) is the ratio of net income to the book value of total assets. GROWTH opportunities are measured by the market value of equity plus the book value of debt divided by the total assets. * , and *** represent statistical significance at the 10%, and 1% levels, respectively. median of RD_TA among firms is 3.9 %, although its average is 14.5%, revealing a widely-distributed and right-skewed R&D expenditure among US firms. The time trend of R&D investment, bank debt, and public debt during the sample period are illustrated in Figure 1.
The average LNSALE is 4.48, meaning that most firms are adequately large. The average cash ratio is 0.14 and growth opportunities reach 2.66, with a few firms exhibiting negative or extremely high growth rates. Other control variables are displayed in Panel B of Table 1.  (1)).
BANKDEBT_TA and PUBLICDEBT_TA are treated as endogenous variables and instru-mented by retained earnings (RE_TA), and firm working capital (WC_TA). See Table 1a for variable definitions. Columns (1) and (2) show the results of the estimated model for BANKDEBT_ TA and PUBLICDEBT_TA using the above instrumental variables and all control variables in equation (1)   Note: * , ** , and *** represent statistical significance at the 10%, 5%, and 1% levels, respectively.
Panel C of Table 1 reports the t-test of independent variables among R&D quintiles. Bank debt and public debt have significant differences between low and high R&D quintiles. This relationship was further checked by running regressions. Panel D reports the Pearson correlation coefficients among all independent variables and confirms low levels of correlations among all of them. Table 3 reports the 2SLS regression results of the use of bank debt and public debt on R&D investment. Columns (1) and (2) show the outcomes of the first stage wherein dependent variables involve BANKDEBT_TA and PUBLICDEBT_TA concurrently. The coefficients of three instrumental variables are statistically significant at the 1% level in both equations. The F value of excluded instruments is very high with a p-value of 0.000, suggesting that the instrumental variables are good predictors. The coefficients of LNSALE and GROWTH on BANKDEBT_TA are positive, suggesting that larger firms and firms with high growth opportunities prefer using more bank loans. Furthermore, the coefficient of SOX2002 is positive and significantly related to PUBLICDEBT_TA at the 1% level, implying that the SOX helps improve financial disclosure and information quality and is good for public debt financing. Table 3 reports the two-stage least squares regression results of firm R&D investment and debt maturity structure, where the ratio of R&D expendi-ture to total assets (RD_TA) is a dependent variable (see equation (1)).

2SLS results
BANKDEBT_TA and PUBLICDEBT_TA are treated as endogenous variables and instrumented by firm retained earnings (RE_TA), and working capital (WC_TA). The first-stage results generated the estimated values of BANKDEBT_TA and PUBLICDEBT_TA as reported in columns (1) and (2). Column (3) reports the second stage of 2SLS estimation results of R&D investment (RD_TA) on bank debt and public debt in equation (1). The test of weak instruments examines the null hypothesis that the instruments are weak at the 5% level of significance. The test of over-identifying restriction examines the null hypothesis that all instruments are valid. The standard errors of estimated coefficients are clustered by the firm and displayed in parentheses. Table 3 reports the second stage of 2SLS results. The coefficient of bank debt on firm R&D investment is positive and significant, implying that banks assist in collecting private information to mitigate information asymmetry and help keep confidentiality and maintain advantages. As a result, lower loan spreads or better nonprice loan terms may be offered. All these cost advantages motivate higher R&D spending. Bank debt use fosters R&D investment, whereas public debt hinders it. These findings are consistent with the idea of Yosha (1995) indicating that the public debt has high costs of information disclosure, which are harmful to firm investments.

Column (3) of
Most of the control variables show the expected signs, except for LNSALE. As shown, older firms or those with more cash or higher growth opportunities tend to invest more in R&D spending. Moreover, the coefficient of SOX2002 is positive and statistically significant, revealing that firms invest more in R&D investment after the SOX was implemented in 2002.  Note: *, **, and *** represent statistical significance at the 10%, 5%, and 1% levels, respectively.

Industry effects
equation (1), bank debt is positively related to firm R&D investment as confirmed in column (1) of Table 4. The coefficient is significant and even higher compared to that in Table 3. Interestingly, the interaction of SOX2002 and public debt over the post-SOX period turns out to be significantly positive, but the interaction of SOX2002 and bank debt becomes negative. These findings imply that while the SOX regulates businesses to disclose accurate and complete information, the advantages of banks disappear. Instead, public debt benefits more and becomes a good debt financing arrangement after SOX.
These results are interpreted as indicative of less information asymmetry between firms and creditors over the post-SOX period. Pre-SOX, public creditors did not prefer to finance firms with high information asymmetry as they were at a disadvantage relative to the banks that could access inside information on borrowing firms (Fama, 1985).
Due to the increase of information transparency post-SOX, public creditors have become willing to provide more funds to innovative firms without too much concern about the disadvantages of the information. Table 4 reports the DID estimation results of firm R&D investment on debt maturity structure, where the ratio of R&D expenditure to total assets (RD_TA) is a dependent variable (see equation (1)).  (1). Column (1) reports the results of the second-stage 2SLS estimation. The effect of the enactment the SOX is further re-examined by adding the interaction terms into equation (1). The results are presented in column (2). ACCELERATED_ FILER is a dummy variable which equals one if it is an accelerated filer to comply to the SOX (firm with market value of equity larger than $75 million), and zero otherwise. SOX2002 is a dummy variable which equals one if the year is after the SOX in 2002, and zero otherwise (see Table 1a for the variable definitions). The standard errors of estimated coefficients are clustered by firm and displayed in parentheses.
To confirm the robustness of the findings, the impact of the enactment of the SOX in 2002 is re-examined to avoid the confounding effects in the results. Accordingly, the DID robustness test is employed. It was first necessary to identify a control group that is not affected by SOX2002. Although the implementation of SOX2002 was mandatory, the Securities and Exchange Commission allowed small (market values of equity under $75 million) firms to have more time to comply with the Act (Abdioglu et al., 2013). Hence, for the control group, any change is not expected between R&D investment and debt structure after the enactment of the SOX. The treatment group is an accelerated filer which includes companies with market value of equity higher than $75 million. A dummy variable (ACCELERATED_ FILER) is created which equals one if a firm is classified as an accelerated filer, and zero otherwise.
It is expected that the relationships among R&D investment, bank debt, and public debt are more pronounced for the treatment group over the post-SOX period. The positive and significant coefficient of SOX2002 ACCELERATED_FILER  BANKDEBT_TA addresses the marginal effect of SOX2002 on the BANKDEBT-RD relation for accelerated filers. Similar results are also found with PUBLICDEBT-RD relation for accelerated filers. The outcomes support the argument that the enactment of the SOX mitigates the information asymmetry. Table 5 reports the DID estimation summary for a time comparison before and after the enactment of the SOX between the accelerated filer and non-accelerated filer groups. For the control group, results support the authors expectation that the signs of bank debt and public debt remain unchanged after the SOX enactment. For the treatment group, the DID results are higher for public debt compared to bank debt in the accelerated filer. Therefore, public creditors obtain more benefit from the enactment of the SOX than the Note: * , ** , and *** represent statistical significance at the 10%, 5%, and 1% levels, respectively.
banks. The DID tests further indicate that following the SOX enactment, the effect of public debt on R&D investment for firms in the accelerated filer increased relatively more than for those in a non-accelerated filer group. Therefore, public debt is significantly negatively related to R&D investment, and this relation is significantly affected by the enactment of the SOX. On the contrary, bank debt has a consistently positive impact on R&D investment, and the magnitudes are robust over the DID model specifications and controls.

CONCLUSION
Using a 2SLS approach and multivariate difference-in-differences analysis for a large sample of US firms, several interesting results are obtained.
A positive relationship exists between R&D investment and bank debt use, but an inverse impact on R&D investment occurs in the presence of public debt. One possible explanation is that although the R&D projects prefer information confidentiality, innovative firms avoid public debt use to reduce high information disclosure cost. Instead, bank debt becomes the better choice due to private information advantages. Moreover, firms with high R&D investment could easily borrow funds from banks with lower spreads or better nonprice loan terms. Bank lending also fosters R&D investment. These findings are robust after conducting the DID tests.
The SOX regulates businesses to disclose accurate and complete information, but the advantages from banks are reduced. Instead, the post-SOX is good for public debt financing. In addition, although the enactment of the SOX significantly mitigates the asymmetric information, such effect is for large firms only, the asymmetry remains unchanged for small corporations.
Finally, it is shown that old firms, firms with more cash, or those with higher growth opportunities invest significantly higher spending on R&D. Overall, the findings shed more light on the effect of debt maturity structure on substantial innovative investment. More importantly, it helps explain the role of corporate debt and bank loans in financing and monitoring corporate R&D investment for long-term economic growth in the US. Note: * , ** , and *** represent statistical significance at the 10%, 5%, and 1% levels, respectively.