Investment Flexibility and Loan Contract Terms

Transcription

1 Investment Flexibility and Loan Contract Terms Viet Cao Department of Accounting and Finance, Monash University Caulfield East, Victoria 3145, Australia Viet Do Department of Accounting and Finance, Monash University Clayton, Victoria 3800, Australia Tram Vu Department of Accounting and Finance, Monash University Caulfield East, Victoria 3145, Australia This version: 4 April 2015 PLEASE DO NOT QUOTE WITHOUT PERMISSION 1

2 Investment Flexibility and Loan Contract Terms Abstract We investigate if borrowers investment flexibility can influence loan contract terms. We test this relationship in the bank loan setting and find that borrower with higher investment flexibility are subjected to higher loan spreads. This effect appears to concentrate among borrowers with potentially heightened agency costs. These borrowers can mitigate the increase loan spread by pledging collateral or accepting stricter loan covenants. We demonstrate that while investment flexibility increase loan spread for opaque borrowers, there are a number of key covenants that they can use to fully mitigate the agency costs of investment flexibility in loan price. Keywords: Loan spreads, loan terms, agency costs, real option, investment flexibility 2

3 1. Introduction There has been a growing debate in the extant literature in economics and finance on the non-trivial impact of investment flexibility on many important aspects in firms investment and financing environments. Often defined as the extent to which firms can change their planned investment and disinvestment patterns (Gross and Khan, 2010), a lack of investment flexibility can distort the efficient allocation of resources at firm level. Caggese (2007) argues that a lack of investment flexibility interacts with financial constraints to amplify the inefficient allocation of firms investment where fixed investment capital is inefficiently high (low) during downturns (upturns). Investment flexibility is also modeled in several contingent claim models as reducing default risk (Mauer and Triantis, 1994; Aivazian and Berkowitz, 1998). Not surprisingly, several studies link investment flexibility with costs of capital. Chirinko and Schaller (2009) argue that the discount rate used by firms with investment flexibility is both economically and statistically lower than that used by firms lacking this flexibility. In Zhang s (2005) and Cooper s (2006) models, the inefficiently high level of fixed capital investments during downturns of firms with a lack of investment flexibility gives rise to their higher systematic risks and a higher required rate of return by their equity holders. Titman et al. (2004) model loan spread as a function of investment flexibility due to both the real option effect and agency costs. While there has been some attempt in empirically verifying the association of investment flexibility with cost of equity capital (Gulen et al., 2008; Docherty et al., 2010; Ortiz-Molina and Phillips, 2014), the evidence on investment flexibility and debt characteristics is sparse. MacKay (2003) empirically studies the relationship between investment flexibility and firms overall financial leverage, maturity 3

4 structure (i.e. long term versus short term debts) and credit structure (i.e. public versus private debts). Our paper aims to investigate how private lenders (i.e. banks) perceive the contribution of firms investment flexibility to default risk and agency costs, and the consequent impact of investment flexibility on loan contracting terms (i.e. loan spreads, maturity, collaterals and covenants). Investment flexibility can potentially affect loan spreads through two channels. First, the real option channel suggests that investment flexibility allows firms to flexibly adjust their capital stocks to respond to macro-economic and product demand shocks. Mauer and Triantis (1994) model that this real option effect lowers firms default risk and expected reorganization and recapitalization costs. This view is also shared by Aivazian and Berkowitz (1998) in their discrete time model. Further, Zhang s (2005) and Cooper s (2006) models, allowing for asymmetric adjustment costs (i.e. expansion is more costly than contraction), conjecture that during downturns, firms with high investment flexibility are not burdened with unproductive idle capital. This view implies that firms with high investment flexibility are less adversely affect by reduced product market demands during downturns and therefore are less risky. Consequently, investment flexibility is potentially negatively associated with loan spreads due to the alleviation of firms systematic risk. On the other hand, Titman et al. (2004) suggest that the real option effect only has implication to debt-holders during downturns when borrowers probability of default is higher. In their model, investment flexibility allows firms to cut back on investments and reduce the quality of assets during downturns, hence lower collateral value. Titman et al. s (2004) model therefore predicts a positive association between 4

5 investment flexibility and loan spreads. Overall, the real option effect on the association between investment flexibility and loan spreads is theoretically ambiguous. A negative association is consistent with viewing firms as going concerns that become less risky due to less excess capacity when the product market demand is low during downturns. A positive association is consistent with the potential deterioration of collaterals that might be called for when default probability is heightened during the same period. 1 In addition to the real option effect, investment flexibility may affect loan spreads by intensifying the agency problem between shareholders and debt-holders. Jensen and Meckling (1976) suggest two mechanisms, i.e. risk shifting and asset substitution, through which the interests of shareholders and debt-holders are misaligned. Several contingent claim models, such as Green and Talmor (1986), Mello and Parsons (1992), Mello et al., (1995) and Leland (1998), conjecture that investment flexibility facilitate risk shifting and asset substitution. Titman et al. s (2004) model suggests that banks, in anticipation of this ex-post behavior, increase the spread they charge firms with high investment flexibility. This positive relation between investment flexibility and loan spread is independent of the real option effect. Beside loan spreads, lenders also rely on alternative non-price mechanisms, for instance, loan maturity, collaterals, and covenants, to align borrowers' interest with that of shareholders. MacKay (2003) suggests that firms with high investment flexibility have high financial leverage as the agency costs can be curbed using contractual terms such as collaterals and covenants. Also, banks are more likely to 1 While the collateral value in liquidation is important in determining loan prices, according to Myers (1977, p. 155), for most lenders, (t)heir loans value depends on the value of the firm as a going concern, not on the value of any specific physical assets. 5

6 demand collateral from borrowers with lower quality to secure the loans (Boot et al., 1991; Holmtrom and Tirole 1997). However, one may argue that firms with investment flexibility may be willing to supply collateral to signal their willingness to respect debt-holders interests, similar to the signaling motivation by better quality borrowers in Besanko and Thakor (1987). It is therefore unclear how investment flexibility is related to collateral. Mauer and Triantis s (1994) model predicts that investment and financial flexibility, defined as the ability to renew or restructure debts at low costs in the event of financial distress, are substitutes. MacKay (2003) reports evidence supporting this view. Similarly, Gamba and Triantis (2008) observe that the negative effects of limited financial flexibility can be mitigated when physical capital is more flexible. Along these lines, firms with high investment flexibility should therefore expect stricter covenants and / or collaterals in their loan contracts. Theories are also divided on how investment flexibility might affect debt maturity. Firms with investment flexibility may wish to signal their genuine intention not to take advantage of the agency problem by accepting short maturity loans, along the lines of Flannery (1986) and Diamond (1991). In addition, short maturities may also be used to curb the agency conflict potential in firms with high investment flexibility. Barnea et al. (1980) argues that risk shifting can be mitigated by shortening term loans as lenders have the option to terminate the loan or reset the loan contract terms at shorter intervals. Along the same vein, Childs et al. s (2005) model, which allows for the joint determination of investment and financing decisions at the firm level, predicts that the use of short term debts can significantly reduce the agency costs of under- and over-investment. 6

7 On the other hand, Mello and Parsons (1992) argue that agency costs may increase as maturity lengthens. Longer maturity debts means that shareholders call option on the firm also has longer maturity, making the option more likely to finish in-the-money. Shareholders are therefore more likely to service the debts to maintain their option, hence the lower probability of default. While MacKay (2003) finds some statistically weak result that firms with high investment flexibility have more long term debts, we seek to reexamine this relation at loan contract level. We find that firms with higher level of investment flexibility face a higher cost of debt on their loan contracts after controlling for borrower characteristics, loan terms, and macroeconomic conditions. This result is consistent with both the agency cost channel and the real option effect along the lines of Titman et al. (2004) where the primary concern is the deterioration of the collaterals of firms with high investment flexibility during downturns. To shed further light on which channel drives our result, we investigate whether the positive association between investment flexibility and loan spreads holds among firms with different degrees of agency problems. Our results show that the relation between investment flexibility and loan spreads holds only among small firms, firms with no S&P credit ratings for senior debts, and firms with high managerial discretionary accruals. According to MacKay (2003), small firms are more prone to the agency problem arising from risk shifting due to more concentrated managerial power and share ownership and growth options. Lacking bond credit ratings and high discretionary accruals (Hutton et al., 2009) both imply an information environment with high asymmetry between firms management and outsiders where agency conflicts are more likely to exist. Our results therefore 7

8 suggest that the agency costs drive the positive association between investment flexibility and loan spreads, consistent with several theoretical models, including Titman et al. (2004). The lack of statistical significance of this relation in firms with potentially low agency costs reflects the division in the theories on the real option effect depending on whether banks view firms as going concerns or focus on the liquidation value of collaterals. We also show that lenders and borrowers are able to negotiate alternative channels to mitigate the effect of investment flexibility on agency costs. Loans to firms with flexible investment are characterized with more collateral and stricter covenants, consistent with the conjecture in MacKay (2003). Furthermore, firms with flexible investment are also more likely to have loans with shorter maturity, consistent with Barnea et al. (1980) and Childs et al. (2005) that shortening maturity can curb the agency costs. Finally, we show that by accepting shorter maturity, more collateral or stricter covenants, firms with high investment flexibility can reduce the high loan spreads charged by banks in anticipation of the potential agency costs due to risk shifting or asset substitution. Our paper makes several important contributions. First, this paper extends the real option literature by providing direct evidence on the interaction between investment flexibility and loan contract terms. While the relation between investment flexibility and loan spreads is theoretically modeled in Titman et al. (2004), we provide the first empirical evidence for this relation. The relation between investment flexibility and non-price debt contract terms including debt maturity, collaterals and covenants is first empirically verified at loan contract level in this study. We therefore bring empirical evidence for several theoretical models advocating this relation. 8

9 We also enrich the literature in cost of capital and debt contracting. While previous empirical literature tends to focus on the impact of investment flexibility on cost of equity capital (Gulen et al., 2008, Docherty et al., 2010; Ortiz-Molina and Phillips, 2014), our study sheds new light of its impact on the cost of capital from the perspective of debt holders. Our investigation of the substitution between price and non-price loan contract terms to curb the potential agency problems arising from investment flexibility also extends our understanding on debt contracting. The remainder of the paper is structured as follows. Section 2 discusses the model used in this study. Section 3 specifies the data sources and variable construction. Section 4 to 6 present the results for the effects of investment flexibility on loan spread and other loan terms. Finally, section 7 concludes the study. 2. The model We adopt the following regression to test whether companies with high investment flexibility pay a higher risk-adjusted loan spread, controlling for firm characteristics, loan characteristics and macroeconomic conditions: AISD = β 0 + β 1 (Flex) + β i (Loan i ) + β j (Borrower j ) + β k (Controls). (1) The variables are defined as follows: AISD: The dependent variable is All-in-spread-drawn (AISD) which represents the interest rate margin over LIBOR on drawn loan amount plus annual fees. This variable is expressed in basis points. Flex: Central to our research question, this variable measures investment flexibility. This concept covers several aspects, from the speed of depreciation to the extent to which assets are rented as opposed to owned, to asset 9

10 redeployability. 2 Motivated by the theory, the extant empirical literature and the practical motive, we use the speed of depreciation to proxy for investment flexibility. Titman et al. s (2004) model on investment flexibility and bank loan spread provides the most direct theoretical backbone for our empirical endeavor. Investment flexibility in Titman et al. (2004) refers to firms ability to adjust the quality level of their assets through the depreciation of existing assets together with new investments. Depreciation is a natural channel through which the capital stock evolves even in the absence of a rental market or a market for second-hand assets. While some studies use asset tangibility as a proxy for investment flexibility (Gulen et al., 2008; Docherty et al., 2010), it does not suit the purpose of proxying for investment flexibility in Titman et al. s (2004) model because it describes the asset or quality level, rather than the change in the quality level. Depreciation is also used in the industrial organization literature to describe investment flexibility (Chirinko and Schaller, 2009; Farinas and Ruano, 2005). Finally, a practical reason to use depreciation ratio is the availability of this variable. A proxy for the redeployability of firms physical assets would result in a much smaller sample as in the case of MacKay (2003) and Ortiz-Molina and Phillips (2014). Our measure of investment flexibility is firm level depreciation ratio, calculated as the annual depreciation expense divided by the beginning of the year net fixed assets. The shorter the useful life of an asset (i.e. the higher depreciation ratio), the easier it is to replace them with new assets, hence the higher level of flexibility. Loan i : vector of loan characteristics including the following variables, 2 MacKay (2003) also classifies labour flexibility, i.e. the ability to hire and fire workers, as part of investment flexibility. Consistent with the industrial organization literature, our focus is on the ability to adjust firms capital stock only. 10

11 LNMAT: Measured as the natural logarithm of loan maturity in number of months. SECURED: A binary variable taking the value of 1 for secured loans and zero for unsecured loans. 3 REVOLVER: A binary variable taking the value of 1 if the loan facility is a revolving facility and zero otherwise. LNLOANSIZE: Measured as the natural logarithm of loan facility amount adjusted for inflation in year 1983 dollars. Borrower j : vector of borrower characteristics including the following variables, LNASSETS: Natural logarithm of borrower s book value of total assets adjusted for inflation in year 1983 dollars. LEVERAGE: Borrower's leverage ratio calculated as book value of total debts divided by book value of total assets. CURRENT: Borrower's current ratio calculated as current assets divided by current liabilities. LNCOVERAGE: Calculated as natural logarithm of (1 + EBITDA/Interest expenses). PROFITABILITY: Ratio of EBITDA over sales. MTB: Borrower's market to book ratio calculated as ratio of (book value of assets book value of equity + market value of equity) to book value of assets. 3 Bharath et al. (2011) and Saunders and Steffen (2011) documented that the secure status as recorded on Dealscan is subject to missing information in several instances. They treated loans with no record of secure status as unsecured loans and conduct robustness on a subsample of loans with recorded secure status. We follow the same approach. 11

12 Controls: vector of control variables including dummies for borrower credit rating (AAA, AA, A, BBB and other ratings), loan purpose dummies, loan year dummies, and borrower industry dummies (based on one-digit SEC codes). We estimate equation (1) using pooled OLS regression. The standard errors are adjusted for heteroskedasticity and clustered at the firm level (Saunders and Steffen, 2011) Data and sample The sample in this study is constructed using three main data sources: 1) Loan Pricing Corporation DealScan (LPC) database and 2) Merged CRSP Compustat database. The LPC database provides information about loan characteristics such as loan price, maturity, collateral, covenants, and loan purpose. Each loan facility is matched with their borrower characteristics obtained from the Merged CRSP Compustat database. For every loan in year t, if the loan active date is six months or more after its firm s Compustat fiscal year ending month, we match it with the Compustat financial information for the same fiscal year. If the loan active date is less than six months after the fiscal year ending month, we match it with the Compustat financial information for the previous fiscal year. This is the same process described in Bharath et al. (2011). 5 Compustat also provides borrowers' primary SIC code. We exclude all loans extended to financial services borrowers (SIC codes between 6000 and 6999). Our final sample includes 10,594 loan facilities with the dollar amount of close to 3.8 trillion US dollars during the period from 1995 to [Insert table 1 here] 4 Our results are robust when clustering at loan deal level. 5 The matching process is aided by the Dealscan-Compustat link file that identifies the GVkey of borrowers in LPC database. We thank Professor Michael R. Roberts for sharing this link file. Details of this link file are described in Chava and Roberts (2008). 12

13 Table 1 provides the descriptive statistics for our data. Panel A shows the calendar year distribution of the loans during our sampling period. There was no clear trend in the number of loans and loan amount from 1995 to The loan number and amount both dropped considerably after the 2008 financial crisis. Panel B shows the main purposes these loans are used for, most common of which are working capital and general corporate purposes. Panel C lists the one-digit SIC code of the borrowers in our sample. The main concentration is among borrowers in the manufacturing sector (SIC code between 2000 and 3999), wholesale and retail sector (SIC code between 5000 and 5999) and service sector (SIC code between 7000 and 7999). Panel D shows the distributions of borrower rating status in our sample. About half of the samples are rated borrower and the other half are unrated. [Insert table 2 here] Table 2 reports summary statistics of the key loan characteristics and borrower characteristics in our sample. The data are winsorized at 1% and 99% levels to remove extreme outliers. The median loan spread (All-in-Spread Drawn, AISD) is 150 bps, median maturity 57 months and median facility size $US150 million (mean facility size $US million). The median book value of assets for our borrowers is $US million (mean assets $US4.2 billion). Both facility size and borrower size are highly skewed, indicating strong heteroskedasticity in our sample. 4. Investment flexibility and loan spread This section discusses the result of model (1) to investigate the effect of investment flexibility on loan spread. Model (1) is estimated using pooled OLS technique. The result is presented in Table 3. Column (1) of Table 3 reports the impact of investment flexibility on loan spreads and includes only loan 13

14 characteristics. Column (2) includes only borrower characteristics, while column (3) presents the full model, using both borrower and loan characteristics as control variables. The result suggest that there firms with higher level of investment flexibility pay higher loan spread. The coefficient in the full model reported in column (3) is positive and significant at 5% level. [Insert table 3 here] The results for control variables are consistent with prior literature on the determinant of loan spread. For loan characteristics, larger loans and revolving loans are associated with lower loan spread while loans with longer maturity, collateral and strict covenant attract higher loan spread. For firm characteristics, larger borrowers, higher rating borrowers and borrowers with higher interest coverage are paying lower loan price. At the same time, borrowers with higher leverage ratio and unrated borrowers are paying higher loan spread. The result that borrowers with higher investment flexibility pay, ceteris paribus, higher loan spreads is consistent with Titman et al. s (2004) agency channel where, in anticipation of potential risk shifting and asset substitution facilitated by investment flexibility, banks charge higher loan spreads. This pattern might also be consistent with the real option effect advocated in Titman et al. (2004) where investment flexibility may lead to the deterioration of the quality of collaterals during downturns. It is, however, inconsistent with the insight in Zhang (2005) and Cooper (2006) where adjustment costs are asymmetric, and firms with high investment flexibility, with less idle physical capital, are less adversely affected by the negative demand shocks. Section 5 further investigates the relative contribution of agency costs and the real option effects. 14

15 5. The Importance of Agency Costs of Debt In this section, we explore the importance of the agency costs of debt to the positive association between investment flexibility and loan spreads. To differentiate this channel with the real option effects, we examine the pattern among firms with potentially different degrees of agency problems. The agency channel suggests that the pattern continues to hold among firms with a high probability of agency problems. The real option effect advocated by Titman et al. (2004) suggests that the pattern continues to hold even in the absence of agency costs, while the real option effect along the lines of Zhang (2005) and Cooper (2006) predicts the opposite relation (i.e. high investment flexibility firms enjoying lower loan spreads). We test these views by running model (1) on different group of borrowers. Table 4 present the result of our tests. [Insert table 4 here] We use a number of well established proxies, including firm size, the availability of S&P senior debt ratings, and managerial discretionary accruals, to capture the severity of agency problems. First, MacKay (2003) suggests that small firms are more likely to have more severe agency problems as these firms are often characterized with more concentrated managerial power, share ownership and growth options. Using total assets as a proxy for size, we report the results for the subsamples of small and large firms in column (1) and (2) (Table 4). Only small borrowers pay higher loan spreads for higher investment flexibility. There is no statistical difference in the loan spreads between more and less flexible firms if they are larger borrowers. In column (3) and (4) (Table 4), we split our sample into borrowers with S&P senior debt rating (rated borrowers) and those without S&P senior debt rating (unrated 15

16 borrowers). Having debt ratings help decrease the information asymmetry between corporate managers and outsiders, hence reducing the potential of risk shifting and asset substitution. Our results show that unrated borrower pay higher loan spreads for their investment flexibility while rated borrowers do not. Finally, we split our sample using an accounting based proxy for information asymmetry proposed by Hutton et al. (2009). The opacity measure is the sum of managerial discretionary accruals in the past three years. Hutton et al. (2009) argue that firms with large managerial discretionary accruals are more likely to manage their reported earnings and therefore report less firm specific information to outsiders. Columns (5) and (6) of Table 4 show that only borrowers with higher level of information opacity pay higher spread for investment flexibility. Overall, the results suggest that the increase in loan spreads caused by investment flexibility of borrower is concentrated only among firms with heightened probability of agency problems where the problem of moral hazard and increasing cost of monitoring is most severe. The results are consistent with the agency channel in Titman et al. (2004). The insignificant relation between investment flexibility and loan spreads in firms with less agency problems is also consistent with the ambiguity in the theories where the real option effect can imply either a positive relation (Titman et al., 2004) or a negative relation (Zhang, 2005; Cooper, 2006). 6. Investment flexibility and non-price loan terms In this section, we ask the question of whether investment flexibility also affect loan terms other than loan spreads. We examine three important loan contract features including loan maturity, collateral and covenants. 16

17 6.1. Loan maturity Loan maturity is driven by both demand and supply sides. Flannery (1986) and Diamond (1991) argue that high quality borrowers demand shorter maturity loans to signal their unobservable credit quality to the market (demand side), while low quality borrowers are only provided with shorter maturity loans due to the high monitoring costs that banks incur (supply side). The results in section 5 suggest that investment flexibility increases the agency costs potentially due to risk shifting and asset substitution, and hence attracts higher loan spreads. Firms with high investment flexibility may be willing to accept short loan maturity to signal their willingness to avoid risk shifting and asset substitution facilitated by their technologies. Alternatively, in anticipation of the agency costs, banks may require short maturities to facilitate their monitoring, along the lines of Barnea et al. (1980) and Childs et al. (2005). On the contrary, Mello and Parsons (1992) argue that longer maturity debts means that shareholders call option on the firm also has longer maturity, making the option more likely to finish in-the-money. Shareholders are therefore more likely to service the debts to maintain their option, hence the lower probability of default and the lower required monitoring efforts. [Insert table 5 here] Table 5 reports the results on the impact of investment flexibility on loan maturity. Column (1) of Table 5 shows the results for the entire sample. Overall, investment flexibility reduces loan maturity on average for the entire sample. The results in columns (2) and (3) for rated and unrated borrowers show that the pattern in the overall sample is driven by unrated borrowers. The coefficient of investment flexibility among unrated borrowers is negative and significant at 1% level while such 17

18 coefficient for rated borrowers is not statistically significant. Qualitatively similar results (available upon request) are found when splitting borrowers by total assets and managerial discretionary accruals. The results support the view that agency costs drive the relation between investment flexibility and loan maturity. Firms with investment flexibility either choose short maturity to signal their genuine intention in the efficient use of the borrowed funds (Flannery, 1986; Diamond, 1991), or are required to do so by banks to help with their monitoring (Barnea et al., 1980; Childs et al., 2005). MacKay (2003) documents a positive (but statistically weak) relation between investment flexibility and the ratio of long term debts to total debts. The author relies on the use of the use of collateral and covenant to mitigate the agency costs that drive the dynamic between investment flexibility and debt maturity. Our results shed new light on this relation by unveiling the relation at loan contract level. Banks appear to rely on short maturity to facilitate their monitoring when the agency costs are potentially acute. We further investigate the roles of collateral and covenant in mitigating agency costs of investment flexibility in the below sections Collateral Similar to loan maturity, collateral can be viewed from both supply and demand sides. On the supply side, firms with investment flexibility may be willing to supply collateral to signal their willingness to respect the bank s interest, mirroring what better quality borrowers would do to signal their credit quality along the lines of Besanko and Thakor (1987). On the other hand, banks are more likely to demand collateral from borrowers with higher agency costs (MacKay, 2003). Firms with 18

19 investment flexibility may also be more willing to accept less flexible financing arrangement (Mauer and Triantis, 1994; MacKay, 2003; Gamba and Triantis, 2008). [Insert table 6 here] Table 6 reports the relation between investment flexibility and loan collateral in general and this relation for rated and unrated borrowers separately. We use dummy variable secured to flag whether a loan is secured by collateral. This dummy variable takes the value of 1 if the loan has collateral and zero otherwise. Bharath et al. (2011) and Saunders and Steffen (2011) documented that the secure status as recorded on Dealscan is subject to missing information in several instances. They treated loans with no record of secure status as unsecured loans and conduct robustness on a subsample of loans with recorded secure status. We follow the same approach. Columns (1) to (3) in Table 6 present the results among the subsample of firms with available data on collateral while columns (4) to (6) use the overall sample and treat loans with missing information as those without collateral. The total observations reduce by a third from 9370 to 6540 when restricting the data to have information on collateral. This reduction is consistent with that reported in Bharath et al. (2011) and Saunders and Steffen (2011). While the investment flexibility coefficient is insignificant in the overall sample, the subsample with collateral data shows more interesting results. The coefficient of investment flexibility becomes statistically significant for the entire subsample with reported collateral data. This pattern is driven by firms with no S&P credit ratings for their senior debts. In untabulated results (available upon request), the 19

20 positive relation between investment flexibility and loan collateral is also driven by small firms and firms with high managerial discretionary accruals. The results indicate that banks require collateral from firms with investment flexibility in general, and particularly so when agency problems are more likely to arise. Again, the signaling motive does not hold. The results also support that firms are more likely to compromise financial flexibility when in possession of investment flexibility, consistent with Mauer and Triantis (1994), MacKay (2003) and Gamba and Triantis (2008) Covenant Covenant restrictions tend to be clustered, i.e. if a firm attracts a particular covenant, it is likely to also attract other types of covenant. Firms with investment flexibility may be willing to accept less flexibility in other areas that loan covenant may impose. Given that firms consider different forms of flexibility as substitutes (Mauer and Triantis, 1994; MacKay, 2003; Gamba and Triantis, 2008) and covenant is another channel that banks may use to curb agency problems (MacKay, 2003), we expect firms with investment flexibility to attract more loan covenant restrictions. In addition to covenant in general, we also focus on those covenants that are more likely to be used to curb agency problems associated with investment flexibility. LPC database reports 24 types of covenants grouped into financial covenants and general covenants, of which many are not directly related to investment flexibility. We isolate four particular covenants that can be linked directly to investment flexibility and test if investment flexibility increases the likelihood of those covenants being imposed to borrowers. Such evidence would lend support to the use of covenants to curb potential agency problems in firms with investment flexibility. 20

21 The four covenants identified are Excess Cash Flow Sweep, Asset Sales Sweep, Debt Issuance Sweep and Equity Issuance Sweep. These are restrictive covenants relating to the use of excess cash, sale of existing assets and the issuance of new debt and equity. These covenants address the possibility of risk shifting and asset substitution that borrowers with high investment flexibility are capable of. We present the results on the relation between investment flexibility and covenant restrictions in table 7. [Insert table 7 here] Column (1) shows the results from the probit regression of the restrictiveness of covenant on investment flexibility. We measure the restrictiveness of covenant following Bradley and Roberts (2004). Specifically, we calculate a covenant index for each loan contract as the sum of six dummy variables representing the existence of a dividend covenant, a debt issuance sweep covenant, an equity issuance sweep covenant, an asset sale sweep covenant, at least two financial covenants, and a secured covenant. A loan is consider to have strict covenant (Strict =1) if the covenant index is greater or equal to 3 and zero otherwise. 6 In column (1) of Table 7, the investment flexibility coefficient is positive and significant at 5% level. This result indicates that overall, firms with higher investment flexibility attract more restrictive covenants. Columns (4) to (7) report the results when the four specific covenants are employed. Investment flexibility appears to increase the requirement for all the four forms of covenants. When Equity Issuance Sweep is used to construct the dependent variable, the investment flexibility coefficient is positive and significant at 10% level. Using the other three specific 6 We also test for the robustness of this proxy by increase the requirement for covenant index to be grater or equal to 4. The results remain robust. 21

22 covenants, the investment flexibility coefficient remains positive and is strongly significant. Next, we explore whether the positive and significant relation between investment flexibility and covenants varies among rated and unrated borrowers. 7 Columns (2) and (3) of Table 7 report that such a relation is present among both rated and unrated borrowers when all covenants are used to measure covenant restrictiveness. Table 8 reports the pattern among rated and unrated borrowers when the four specific covenants are used. Among unrated borrowers, the investment flexibility coefficient is positive and significant when covenant restrictiveness is determined using three out of the four specific types of covenants. 8 Yet, none of the flexibility coefficients is significant among rated borrowers. Overall, the results are consistent with banks using covenants to curb agency problems associated with investment flexibility, and firms with this type of flexibility being willing to accept restrictions in other aspects of their investment and financing environments. The results also lend supports to a conjecture not empirically verified in MacKay (2003) that investment flexibility does not stop firms from accessing debts as covenants might be used to deal with the associated agency costs. [Insert table 8 here] 7. The substitution effect between price and non-price loan terms So far we have established that borrowers with flexible investment pay higher loan prices, have shorter maturity, are more likely to pledge collaterals and accept more restrictive covenants. MacKay (2003) argues that firms with investment 7 The results when splitting firms by the other proxies for the severity of agency problems (i.e. size and managerial discretionary accruals) are qualitatively similar and are available upon request. 8 The only exception is when covenant restrictiveness is determined using Debt Issuance Sweep. 22

23 flexibility have more debts in their capital structure as contractual terms such as covenants and collateral can mitigate risk shifting and asset substitution behaviors. In this session, we explore whether banks accept some level of substitution of these requirements in curbing potential agency problems associated with borrowers investment flexibility. While MacKay (2003) focuses on the trade-off with the amount of credit available to borrowers at firm level, our focus is on the trade-off with loan prices at loan contract level. Arguably these are the most important aspects of firms accessibility to credit. We begin our examination by interacting three other loan terms: maturity (Mat), secured dummy (Secured) and covenant restriction (strict) with investment flexibility. In the regression of loan spreads on independent variables, the coefficients of the interaction terms provide evidence on any potential substitution effect between price and non-price loan terms. The results are presented in table 9. [Insert table 9 here] The coefficient attached to the interaction term between flexibility and maturity is not statistically significant, implying that by accepting shortened loan maturity, borrowers with high investment flexibility may not be able to reduce the loan spread. The coefficients attached to the interaction terms with both collateral and covenants are negative and significant at 5% level. These results indicate that overall, borrowers with high investment flexibility can lower the cost of their borrowing by pledging collateral and accepting more restrictive covenants. Our results lend support to MacKay s (2003) conjecture that collateral and covenants might have helped banks address the agency costs associated with investment flexibility and allow firms with this type of technology enjoy access to credit. 23

24 [Insert table 10 here] Finally, we test the robustness of the substitution effect by exploring whether the four specific covenants identified in section 6.3 can reduce the borrowing cost among firms with more severe agency problems. In Table 10, we restrict our sample to unrated borrowers who, according to our results so far, are more likely to face tighten the loan contract terms in response to investment flexibility. Of these firms, we identify those that accept one of the four specific covenants identified in section 6.3 and those that do not. We expect the investment flexibility coefficient to be less significant among firms that accept one of the four specific covenants. The results are overwhelmingly consistent. The investment flexibility coefficient is strongly significant in columns (2), (4), (6) and (8) where unrated borrowers do not accept one of the four covenant restrictions. However, among unrated borrowers who are willing to accept any of the restrictive covenants, the flexibility coefficients in all four columns (1), (3), (5) and (7) are not statistically significant. The results imply that investment flexibility does not increase loan spreads for unrated borrowers if they are willing to accept any of the covenants that are specifically designed to curb the agency problems associated with investment flexibility. The results support our view on the substitution effect of non-price terms for prices reported in Table 9, and highlight a particular mechanism, i.e. specifically designed covenants, that firms accept to alleviate the pricing impact of the perceived agency costs that their technologies present. 8. Conclusion Using a sample of 10,594 loan facilities with the dollar amount of close to 3.8 trillion US dollars during the period from 1995 to 2010 in the U.S. market, this study 24

25 examines the question of how investment flexibility affects loan contract terms. Consistent with Titman et al. s (2004) theoretical model, we find evidence supporting the view that banks require higher loan spreads to compensate for the heightened monitoring costs to address the agency problems including risk shifting and asset substitution associated with investment flexibility. We do not find evidence supporting the real option effect of this type of flexibility on loan spreads, which is also consistent with the ambiguity from the guiding theories (Titman et al., 2004 versus Zhang, 2005 and Cooper, 2006). We also document the use of non-price loan terms, including maturity, collateral and covenant, for the same purpose. Finally, we also document a substitution effect among the loan terms to address agency costs. Firms with investment flexibility and willingness to pledge collateral or accept strict covenants can reduce the heightened loan spreads charged by banks while shortening loan maturity alone is insufficient. These results are consistent with the view of MacKay (2003) that the presence of collateral and covenants in debt contracts can address the agency costs perceived by debt holders that could have limited borrowers from accessing the credit market. In particular, we find strong evidence that accepting the specifically designed covenants to address the agency problems associated with investment flexibility can completely alleviate the pricing impact of investment flexibility among the borrowers with potentially more severe agency problems. Our results have important implications to both borrowers and banks. Borrowers can confidently have the freedom to employ the technologies that best suit their needs, knowing that pledging collaterals and / or accepting certain restrictive covenants will alleviate the impact on costs of debt of potential risk shifting and asset 25

26 substitution that their technologies might be associated with. This solution is also available to firms with severe perceived agency costs, including small firms and firms operating in an asymmetric information environment. Banks can also be confident that their monitoring costs on firms with high perceived risk of agency problems due to borrowers technologies are acknowledged and reflected in a variety of loan terms which can be substituted and tailored to the need of their borrowers. Given that firms may be limited to the choice of technologies, this awareness can help improve corporate access to bank loans and the efficient allocation of capital. 26

27 References Aivazian, V. A., and M. K. Berkowitz, Ex Post Production Flexibility, Asset Specificity, and Financial Structure, Journal of Accounting, Auditing and Finance, 1998 (13), Barnea, A., R. Haugen, and L. W. Senbet, A Rationale for Debt Maturity Structure and Call Provisions in the Agency Theoretic Framework, Journal of Finance, 1980 (35), Besanko, D. and A. Thakor, A., Competitive Equilibrium in the Credit Market under Asymmetric Information, Journal of Economic Theory, 1987(42), Bharath, S. T., D. Sandeep, A. Saunders, and A. Srinivasan, "Lending Relationships and Loan Contract Terms, Review of Financial Studies, 2011(24), Caggese, A., "Financing Constraints, Irreversibility, and Investment dynamics", Journal of Monetary Economics, 54(2007), Chava, S., and M.R. Roberts, M. R., How Does Financing Impact Investment? The Role of Debt Covenants, Journal of Finance, 2008(63), Chirinko, R. and H. Schaller, The Irreversibility Premium, Journal of Monetary Economics, 56(2009), Cooper, I., "Asset Pricing Implications of Nonconvex Adjustment Costs and Irreversibility of Investment", Journal of Finance, 61(2006), Docherty, P.A., H. Chan, and S. Easton, "Tangibility and Investment Irreversibility in Asset Pricing", Accounting and Finance, 50(2010),

28 Farinas, J.C. and S. Ruano, "Firm Productivity, Heterogeneity, Sunk Costs and Market Selection", International Journal of Industrial Organization, 23(2005), Flannery, M., Asymmetric Information and Risky Debt Maturity Choice, Journal of Finance, 1986(41), Groth, C. and H. Khan, Investment Adjustment Cost An Empirical Assessment, Journal of Money, Credit and Banking, 42 (2010), Gulen, H., Y. Xing, and L. Zhang, "Value versus Growth: Time-varying Expected Stock Returns", Working Paper No (2008), Ross School of Business. Leland, H.E., Agency Costs, Risk Management and Capital Structure, Journal of Finance, 53(1998), MacKay, P., Real Flexibility and Financial Structure: An Empirical Analysis, Review of Financial Studies, 16(2003), Mauer, D. C., and A. J. Triantis, Interactions of Corporate Financing and Investment Decisions: A Dynamic Framework,' Journal of Finance, 49(1994), Mello, A. S., and J. E. Parsons, Measuring the Agency Cost of Debt, Journal of Finance, 47(1992), Mello, A. S., J. E. Parsons, and A. J. Triantis, An Integrated Model of Multinational Flexibility and Financial Hedging, Journal of International Economics, 39(1995),

29 Myers, S.C., Determinants of Corporate Borrowing, Journal of Financial Economics, 1977(5), Ortiz-Molina, H. and G.M. Phillips, Real Asset Illiquidity and the Cost of Capital, Journal of Financial and Quantitative Analysis, (2014), forthcoming. Saunders, A. and S. Steffen, The Costs of Being Private: Evidence from the Loan Market, Review of Financial Studies, 2011(24), Titman, S., S. Tompaidis, and S. Tsyplakov, Market Imperfections, Investment Flexibility, and Default Spread, Journal of Finance, 59(2004), Zhang, L., "The Value Premium", Journal of Finance, 60(2005),

30 Table 1 Distribution of loan facilities Panel A: Number and dollar amount Panel C: Number of facilities by purpose (USD mil) of facilities by year ,881 Refinancing 1, ,920 Working capital 1, ,574 Corporate purposes 2, ,939 Acquisitions & LBO ,680 Other purposes 2, ,480 Total 10, , , , , , , , , , ,305 Total 10,594 3,771,669 Panel C: Number of facilities by borrower SIC code Panel D: Number of facilities by borrower debt ratings SIC=0 20 AAA 87 SIC=1 842 AA 215 SIC=2 2,321 A 1,040 SIC=3 3,035 BBB 1,669 SIC=4 795 Below BBB 2,870 SIC=5 1,737 Non-rated 4,713 SIC=7 1,231 Total 10,594 SIC=8 587 SIC=9 26 Total 10,594 30

31 Table 2 Descriptive statistics for key loan terms and borrower characteristics Variable Obs. Mean Median Std. Dev. Min Max Loan characteristics All-in-spread drawn (bps) 10, Facility amount (USD mil) 10, ,000 Maturity (months) 10, Secured dummy 10, Revolver dummy 10, Borrower investment flexibility Depreciation ratio 10, Other borrower characteristics Total assets (USD mil) 10,594 4, , ,342 Leverage 10, Current ratio 10, Interest coverage 10, Profitability 10, Market to book ratio 10,

32 Table 3 OLS regression of loan yield spreads on investment flexibility This table presents the OLS regression output for All-in-Spread Drawn (AISD) on investment flexibility (LNDEP, i.e. the natural logarithm of the depreciation ratio). Other determinants include loan characteristics (column 1), borrower characteristics (column 2), and both loan and borrower characteristics (column 3). All variables are defined in section 4.2. The standard errors are corrected for clustering at the firm level. ***, **, * represent significance at 1%, 5%, and 10% level, respectively. Dep. Var. = All-in-Spread Drawn (AISD) (1) (2) (3) LNDEP *** ** (15.958) (18.241) (18.145) LNASSETS *** *** (1.579) (1.807) LEVERAGE *** *** (13.042) (10.934) CURRENT (1.761) (1.408) LNCOVERAGE *** *** (2.474) (2.041) PROFITABILITY (15.496) (13.577) MTB *** *** (1.717) (1.431) AAA (36.129) (23.551) AA *** *** (9.232) (6.865) A *** *** (6.558) (5.493) BBB *** *** (5.343) (4.454) OTHERRATINGS *** *** (4.631) (4.044) LNLOANSIZE *** *** (0.922) (1.626) LNMAT *** *** (1.959) (2.061) SECURED *** *** (2.880) (3.129) STRICT *** *** (3.425) (3.321) REVOLVER *** *** (2.044) (2.000) Constant *** *** *** (45.083) (38.384) (42.936) Year dummies YES YES YES Industry dummies NO YES YES Loan purpose dummies YES NO YES Observations 10,086 9,746 9,338 Adj R-squared