BANK LENDING AND INTEREST- RATE DERIVATIVES

Using recent data that cover a full business cycle, this paper documents a direct relationship between interest-rate derivative usage by U.S. banks and growth in their commercial and industrial (C&I) loan portfolios. This positive association holds for interestrate options contracts, forward contracts, and futures contracts. This result is consistent with the implication of Diamond’s model (1984) that predicts that a bank’s use of derivatives permits better management of systematic risk exposure, thereby lowering the cost of delegated monitoring, and generates net benefits of intermediation services. The paper’s sample consists of all FDIC-insured commercial banks between 1996 and 2004 having total assets greater than $300 million and having a portfolio of C&I loans. The main results remain after a robustness check. JEL Classification: G21; G28


Introduction
The relationship between the use of derivatives and lending activity has been studied in recent years. Brewer, Minton, and Moser (2000) evaluate an equation relating the determinants of Commercial and Industrial (C&I) lending and the impact of derivatives on C&I loan lending activity. They document a positive relationship between C&I loan growth and the use of derivatives over a sample period from 1985 to 1992. They find that the derivative markets allow banks to increase lending activities at a greater rate than the banks would have otherwise. Brewer, Jackson, and Moser (2001) examine the major differences in the financial characteristics of banking organizations that use derivatives relative to those that do not. They find that banks that use derivatives grow their business-loan portfolio faster than banks that do not use derivatives. Purnanandam (2004) also reports that the derivative users make more C&I loans than non-users. There are two major research questions that arise in the literature: Does the use of derivatives facilitate loan growth? If not, is there a negative association between lending activity and derivative usage? Using recent data that cover a full business cycle, this study revisits these questions to ascertain whether a direct relationship still exists.
This study differs from the previous research in several aspects. First, it uses more recent data. Few of the previous research studies cover the period from 1996 through 2004.
During this period, the use of interest-rate derivatives for individual banks is even more extensive than in earlier studies, rising from notional amounts of $27.88 trillion at the end of December 1996 to $62.78 trillion at the end of 2004. 1 Given the substantial change in the use 1 The notional amount is the predetermined dollar principal on which the exchanged interest payments are based. The notional amounts of derivatives reported are not an accurate measure of derivative use because of reporting practices that tend to overstate the actual positions held by banks. Even though notional values do of derivatives, the research inferences drawn in the previous studies based on less derivative usage may not hold under the current circumstances. Therefore, the use of more recent data in this study will shed more light on the most recent impact of derivative usage on bank lending activity.
Second, the sample period in this study covers a full business cycle, thereby providing a better indication of the relative variability of lending activities experienced by commercial banks over this period. Brewer, Minton, and Moser (2000) document a universal downward trend of C&I lending over a sample period of 1985 to 1992, a period during which the economy experienced a significant cyclical downturn. In contrast, our sample enables me to focus on a more comprehensive picture regarding the impact of derivative usage on lending activity through the different stages of the business cycle, such as economic boom and economic recession.
Finally, the definitions of several variables in the Call Reports are different prior to 1995. For example, futures and forwards are reported together in the Call Report data. It is more difficult for researchers to examine the effect of different derivative instruments on a bank's lending activities, since swaps and forwards may have different characteristics from futures and options. The sample period of the research in this paper is a time period over which there is a specific definition and consistent measurement of each interest-rate derivative instrument in the Call Reports. Therefore, the construction of these variables will be more accurate and much more detailed than the ones used in previous studies.
The sample in this study represents FDIC-insured commercial banks with total assets greater than $300 million as of March 1996 that have a portfolio of C&I loans. Following not reflect the market value of the contracts, they are the best proxy available for the usage and the extent of usage of interest-rate derivatives. Brewer, Minton, and Moser (2000), we evaluate an equation relating the determinants of C&I lending and the impact of derivatives on C&I lending activity. The major finding in this study is that the interest-rate derivatives allow commercial banks to lessen their systematic exposure to changes in interest rates, which enables banks to increase their lending activities without increasing the total risk level faced by the banks. This consequently increases the banks' abilities to provide more intermediation services. Furthermore, a positive and significant association between lending and derivative activity indicates that the net effect of derivative use on C&I lending is complementary. That is, the complementary effect dominates any substitution effect.
Additionally, this positive association holds for interest-rate options contracts, forward contracts, and futures contracts, suggesting that banks using any form of these contracts, on average, experience significantly higher growth in their C&I loan portfolios.
Furthermore, C&I loan growth is positively related to capital ratio and negatively related to C&I loan charge-offs. The findings in this study are confirmed after a robustness check.
Examining the relationship between the C&I loan growth and derivative usage poses a potential endogeneity problem because the derivative-use decision and lending choices may be made simultaneously. To address this problem, an instrumental-variable approach is employed. Specifically, we estimate the probability that a bank will use derivatives in the first-stage specification, then we use the estimated probability of derivative usage as an instrument for derivative activity in the second-stage C&I loan growth equation. The probit specification for this instrumental variable is based on Kim and Koppenhaver (1992). This paper is organized as follows: The following section describes the sample and data sources. A discussion of the empirical specifications for commercial and industrial lending is provided in the third section. Next, the empirical results are presented in the fourth section. The fifth section provides robustness test results, and the final section concludes the paper.

Data and Sample Description
This section describes the sample selection criteria, the lending activity experience by FDIC-insured commercial banks from the fourth quarter of 1996 through the fourth quarter of 2004, as well as the interest-rate derivative products used by sample banks during the nine-year sample period.

Sample Description
The sample of banks includes FDIC-insured commercial banks with total assets greater than $300 million as of March 1996. Of these institutions, banks that have no commercial and industrial loans are excluded. The sample ranges from 942 banks in March of 1996 to 467 banks in December of 2004. Institutions that are liquidated during the sample period are included in the sample before liquidation and excluded from the sample for the periods after liquidation. Banks that merge during the sample period are included in the sample. By construction, the sample is therefore free from survivor bias. Balance sheet data and interest-rate derivative-usage information are obtained from the Reports of Condition and Income (Call Report) filed with the Federal Reserve System. State employment data are obtained from the U.S. Department of Labor, Bureau of Labor Statistics.

Lending Activity
Because the accessibility of credit depends importantly on banks' roles as financial intermediaries, loan growth is an important measure of intermediaries' activities. Following Brewer, Minton, and Moser (2000), we use C&I loan growth as a measure of lending activity because such a measure performs a critical function in channeling funds between the financial and the productive sectors of the economy. Table 1 presents year-end data for bank C&I loan lending activity for the sample banks from 1996 through 2004. The sample period covers a full business cycle and thereby provides a better indication of the relative variability of lending activities experienced by the commercial banks in different stages of a business cycle. Panels B through E report data for four categories of institutions classified by total asset size. Corresponding to the acceleration of C&I loans in the late 1990s, the average ratio of C&I loans to total assets increases steadily, from 12.44 percent at the year-end of 1997 to 13.15 percent at the year-end of 2000. Then, from year-end 2001 to year-end 2003, the average ratio of C&I loans to total assets exhibits a downward trend, which corresponds to the economic recession beginning in March of 2001. As panels B through E report, this pattern exists across different sizes of banks, with the largest decline occurring for banks having total assets greater than $10 billion. This decline stops at year-end 2004 when the overall economy experiences more rapid growth.

Interest-rate Derivative Products
The use of interest-rate derivatives by banks has grown dramatically in recent years, rising from notional amounts of $27.88 trillion at the end of 1996 to $62.78 trillion at the end of 2004. Four main categories of interest-rate derivative instruments are examined: swaps, options, forwards, and futures. Table 2 presents the notional principal amounts outstanding and the frequency of use of each type of interest-rate derivative by banks from year-end 1996 through year-end 2004. As in Table 1, data are reported for the entire sample of banks and for four subgroups of banks categorized by total asset size. Consistent with the dramatic increase in the use of derivatives in recent years, Table 2 shows extensive participation of banks in the interest-rate derivative markets over the nine-year sample period. Furthermore, the rapid growth in the use of various types of derivative instruments has not been confined to large commercial banks; medium-size and small-size banks have also experienced a tremendous increase in the participation of derivative markets. Table 2, during the entire sample period, the most widely used interestrate derivative instrument is the swap. At the end of 1996, 31.6 percent of banks report using interest-rate swaps. By the end of 2004, the percentage using swaps rise to 37.3 percent. Over the nine-year sample period, more than 95 percent of banks with total assets exceeding $10 billion report using interest-rate swaps.

As shown in
Another notable increase occurred in the forward-rate agreement (FRA) usage. FRA is a contract that determines the rate of interest, or currency exchange rate, to be paid or received on an obligation beginning at some future date. At the end of 1996, 9.02 percent of the sample banks report using FRAs. By the end of 2004, the percentage using FRAs more than doubled. While the percentage of banks participating in the swaps and forwards increased over the sample period, the proportion of banks using interest-rate options fell.
This decline is most notable between year-end 2000 and year-end 2004. With the exception of banks with total assets greater than $10 billion, less than 7.5 percent of banks report having open positions in interest-rate futures.
Finally, less than 3 percent of the sample banks report having open positions in interest-rate swaps, interest-rate options, interest-rate forwards, and interest-rate futures. In contrast, nearly half of the banks with total assets greater than $10 billion report having positions in all four types of interest-rate derivative instruments. This result strongly suggests that large banking organizations are much more likely than small banking organization to use derivatives. As shown in Panel E of Table 2, approximately 25 of the largest banks heavily participated in the interest-rate derivative market, a result similar to the finding of Carter and Sinkey (1998).

Specifications of Variables
Based on the literature regarding the determinants of bank lending, this section describes the specification for intermediation, the independent variables used in the empirical model, and the measure of derivative activities.

The Specification for Intermediation
The foundation of the empirical analysis in this article is the specification for bank lending by Sharpe and Acharya (1992). They regress a measure of lending activity on a set of possible supply and demand factors ( , 1 )

Independent variables (Traditional Supply and Demand Factors)
The explanatory variables represent both supply and demand factors Sharpe and Acharya (1992), among others, 2 relate overall loan growth to capital requirements.
In addition, Sharpe (1995) finds that there is a positive association between bank capital and loan growth. In a more recent work, Beatty and Gron (2001) document that, consistent with Sharpe's finding, banks with higher capital growth relative to assets experience greater increases in their loan portfolios, and banks with weak capital positions are less able to increase their loan portfolios due to capital constraints. When a bank's capital falls short of the required amount, the bank could attempt to raise the capital-to-asset ratio by reducing its assets (the denominator of the ratio) rather than raising capital (the numerator of the ratio).
One way of doing this is to shift the asset portfolio away from lending, such as cutting back its investment in C&I loans. Banks may choose this strategy over equity issuance simply because issuing equity is costly. 3 Therefore, undercapitalized banks are less able to increase their loan portfolios while satisfying the regulatory capital requirements. In contrast, banks with stronger capital positions have more room to expand their loan portfolios and still be able to satisfy the regulatory requirement for the capital-to-asset ratio. If capital-constrained banks adjust their lending to meet some predetermined target capital-to-asset ratios, one would expect a positive relationship between a bank's capital-to-asset ratio and C&I loan growth. In order to control for the effect of capital requirements on C&I lending activity, a measure of the bank's capital-to-asset ratio (CARATIO) is included in the empirical specification for C&I loan growth. CARATIO is measured as the ratio of a bank's total equity capital to total assets at time t-1.
Another factor found to affect loan growth is the quality of a bank's loan portfolio.
Following Sharpe and Acharya (1992), we use C&I loan charge-offs (CILCOFA) as a proxy for loan quality. 4 The variable CILCOFA is constructed as the ratio of C&I loan charge-offs in the last period (t-1) to total assets in the last period (t-1). Charge-offs usually rise during a recession and decline only after an economic recovery. Therefore, a low charge-offs ratio can also be a signal of a favorable economic environment in a bank's geographic region of operations. In addition, the ratio of C&I loan charge-offs to total assets could capture the impact of regulatory pressures on loan growth because regulators often apply pressure to banks to increase their rates of charge-offs. For example, capital-constrained banks may be required to increase their rates of charge-offs so that they can clear the regulatory hurdle for capital ratios by eliminating some of their assets. 5 Therefore, the ratio of C&I loan chargeoffs to total assets could reflect the impact of regulatory pressures on banks' capital management. Each of these reasons suggests that those banks with high charge-offs should, other things being equal, be viewed as less well capitalized than banks with low charge-offs, and are therefore less able to increase their loan portfolios due to capital constraints. For these reasons, one would expect CILCOFA to have a negative association with C&I loan growth.
The relationship between bank health and regional economic conditions is another factor to consider. The idea that regional economic performance affects bank health is intuitive and broadly consistent with the aggregate banking data. 6 Avery and Gordy (1998) find that one-half of the change in bank loan performance from 1984 to 1995 can be explained with a group of state-level economic variables. Also, Bernanke and Lown (1991) and Williams-Stanton (1996) point out that regional economic conditions should influence bank C&I loan growth. The intuition is that banks in states with weak economic conditions are likely to have fewer profitable opportunities than banks in states with stronger economies.
The state employment growth rate ( , 1 j t EMPG − ) is included in the model as a proxy for local economic conditions that are not captured by the other explanatory variables. 7 If state employment growth is a proxy for economic conditions, one would expect EMPG to be positively related to C&I loan growth, other things being equal.

Measure of Derivative Activities
In order to capture the effects of derivative usage on bank-loan growth, we include various measures of participation in interest-rate derivative markets ( , j t DERIV ) in the C&I loan growth specification (the construction of this variable is presented in equation 2). The coefficient estimate on DERIV reflects the impact of derivative usage conditional on adequately incorporating the intermediating process in the remaining terms of the 6 For example, Daly, Krainer, and Lopez (2003) show that there is a significant trackable link between regional economic performance and bank health. Also, Berger, Bonime, Covitz, and Hancock (2000) document that aggregate state-level and regional-level variables are important contributors to the persistence in firm-level performance (i.e., return on assets) observed in the U.S. banking industry. specification. Modern theories of the intermediary role of banks describe how derivative contracting and lending can be complementary activities. Diamond (1984) develops a theory of financial intermediation. In his model, banks optimally offer debt contracts to "depositors" and accept debt contracts from "entrepreneurs." Depositors delegate monitoring activities to banks that have the ability to economize the costs of monitoring loan contracts made with entrepreneurs. However, banks face an incentive problem that originates from the cost of delegated monitoring on behalf of their depositors. Diamond shows that diversification within a bank lowers the cost of delegated monitoring. An implication of his model is that banks should not assume any nondiversifiable risks unless they have special advantages in managing them. Thus in his model, banks find it optimal to hedge all interest-rate risk by interest-rate derivatives. 8 However, even after diversifying, banks may still face systematic risks that cannot be diversified away.
Diamond demonstrates that derivative contracts can serve as a third form of contracting, which enables banks to reduce their exposure to systematic risk in their loan portfolios. This use of derivative contracts to hedge systematic risks enables banks to obtain further reductions in delegation costs, and, in turn, allows banks to intermediate more effectively. Empirically, Brewer, Jackson, and Moser (1996) find that there is a negative correlation between risk and derivative usage for savings and loan institutions. In fact, they find that S&Ls that use derivatives experience relatively greater growth in their fixed-rate mortgage portfolios. Brewer, Minton, and Moser (2000) examine the relationship between lending and derivative usage for an earlier sample of FDIC-insured commercial banks. Their results indicate that banks using interest-rate derivatives, on average, experience significantly higher growth in their C&I loan portfolios. These results are consistent with the notion that derivative usage would help banks better cope with interest-rate risk and thereby enable them to hold more loans to earn more income from their lending activity. If interest-rate derivative activity complements the lending activity as predicted by Diamond's (1984) (2000) also document a similar pattern regarding C&I lending over a sample period from 1985 to 1992, a period during which the economy experienced a significant cyclical downturn. They argue that the downward trend in lending activity and the concurrent increase in the use of interest-rate derivatives suggest that derivative usage might be a substitute for lending activity. They suggest that a negative relationship between derivative usage and lending activity could arise in two cases. The first case is when banks use derivatives for speculative purposes. Gain from speculating on interest-rate changes would enhance revenues from bank trading desks.
The second instance is when banks charge a fee as over-the-counter dealers for placing derivative positions. Pursuit of either of these activities as a replacement for the traditional lending activities of banks would imply that derivative activity would be a substitute for lending activity. If these activities were substitutes, one would expect a negative coefficient on the DERIV variable.
From the above discussion, a specification for Equation (2) can be written as follows: In Equation (2), , j t CILGA is measured as the quarterly change in C&I loans relative to last period's total assets. t D is a time-indicator variable equal to one for period t, or zero otherwise. The variable , 1 j t CARATIO − is the ratio of a bank's total equity capital to total assets in the previous period (t-1).
, 1 j t CILCOFA − is the ratio of C&I loan charge-offs in the previous period (t-1) to total assets in the previous period (t-1).
, 1 j t EMPG − is the state employment growth rate relative to last period (t-1), where EMP equals total employment in the state in which the bank's headquarters are located. The variable , j t DERIV is a measure of participation in interest-rate derivative markets. Table 3 reports summary statistics for the variables used in the estimation of Equation (2). The mean of quarter-to-quarter changes in C&I loans scaled by values of beginning-ofquarter total assets is 0.4 percent over the full sample period. During this period, the average capital-to-asset ratio is 9.45 percent, the average C&I loan charge-offs over assets is 0.05 percent, and the average state employment growth rate is 0.45 percent. Consistent with the data presented in Table 2, 20.78 percent of the sample banks reported using interest-rate swaps during the sample period, 11.26 percent of the sample banks reported using interestrate options, and 8.61 percent reported using FRAs. Only 3.28 percent of the sample banks reported using interest-rate futures. Finally, over-the-counter dealers and subsidiaries of foreign banks comprise only 1.2 percent and 4.5 percent, respectively, of the sample bank observations.

Instrumental Variable
Examining the relationship between the C&I loan growth and derivative usage poses a potential endogeneity problem because the derivative-use decision and lending choices may be made simultaneously. As the data show, the decisions could be made jointly since a bank's C&I lending activity might affect its decision to use derivatives. To address this problem, an instrumental-variable approach is used.
The probit specification for the instrumental variable is based on Kim and Koppenhaver (1992). 9 This probit specification includes the log of bank assets, the capitalto-asset ratio, net interest margin, and the first lag of the dependent variable. Commercial bank size as measured by the logarithm of its total assets is included to control for the differences in derivative use that might be caused by differences in the types of businesses and customers at large and small banks. 10 The capital-to-asset ratio is included in the probit specification because a bank's capital adequacy is a necessary condition for its participation in the derivative market. A bank's net interest margin enters into the equation because banks can use derivatives to lock-in the spread between interest income and interest expense. Since derivative use at time t is usually dependent on derivative use at time t-1, the first lag of the dependent variable is included to take into account the dependence over time. To determine the probability of a bank's derivative usage, the above probit specification for each sample date t is estimated, and then the estimated probability from the first-stage estimation is used as an instrument for derivative activity in the second-stage estimation. 11 The results of this first-stage regression are presented in Appendix A. Overall, the probit results show that, as predicted, bank size, capital-to-asset ratio, and the lagged dependent variable play a significant role in determining the probability of derivative usage by U.S. commercial banks. 9 Brewer, Minton, and Moser (2000) use a similar probit specification in their study. 10 Previous literature finds that size is an important indicator in a bank's derivative activities; e.g., Sinkey and Carter (1997), Kim and Koppenhaver (1992), and Gunther and Siems (1996). 11 A Hausman test indicates that the instrumental variable is a valid instrument.
Since banks' use of derivatives increases during the sample period, a pooled crosssectional time-series regression is employed to incorporate this dynamic effect. 12 Specifically, we run a cross-sectional Ordinary Least Squares regression with C&I loan growth as the dependent variable and then report the time-series means of the parameter estimates and their corresponding t-statistics. The t-values are computed using Newey-West heteroskedasticityand-autocorrelation-consistent errors. We use the predicted derivative use, obtained from the probit specification, to instrument the actual derivative-use variable as an independent variable.

Empirical Results
Using the quarterly change in C&I loans relative to last period's total assets as the dependent variable, we utilize Equation (2) to examine the determinants of C&I lending and the impact of derivatives on C&I lending activity. Table 4 reports the results of pooled crosssectional time-series regressions using quarterly data from March 1996 through December

2004.
Regression (1) of Table 4 is the reduced form of the supply equation that examines the impact of fundamental factors on C&I lending activity. This regression serves as a base for examining the relationship between derivative activity and C&I lending. In regression (1), C&I loan growth is significantly and positively related to the beginning-of-period CARATIO.
This result is consistent with the hypothesis that capital-constrained banks adjust their loan portfolios in subsequent periods to meet some predetermined target capital-to-asset ratios.
Similar to Brewer, Minton, and Moser (2000), we also find a significant, negative association between CILCOFA and C&I loan growth. This negative relationship is consistent with the notion that the charge-offs variable captures the impact of regulatory pressures, a strong economic environment, or both. The previous period's state employment growth variable EMPG fails to enter the equation significantly. This result is inconsistent with Brewer, Minton, and Moser (2000), who study a sample of banks that predates the advent of interstate banking. During the course of the 1990s, deregulation in the banking industry led to consolidation and to banks' geographic expansion. As a result, U.S. banks have also become more geographically diversified. In fact, the regression results might suggest that state economies play a lesser role in affecting banks' health and performance following the full expansion of interstate banking.

Regressions (2) and (3) include different measures of derivative activity. Regression
(2) augments the predicted probability of derivative usage in any type of interest-rate derivative contract (DERIV). Regression (3) decomposes the DERIV variable into four types of interest-rate derivative instruments: SWAPS, OPTIONS, FORWARDS, and FUTURES.
Each type of derivative activity is estimated using the probit specification discussed earlier in this section. The estimates generated in the probit specification are then used in conjunction with the supply and demand factors in the second-stage regression to predict C&I loan growth.
Columns (2) and (3) of Table 4 report the estimation results for the derivativeaugmented regressions. First, the coefficient estimates on CARATIO, CILCOFA, and EMPG are qualitatively similar to those in the base model. Second, the CARATIO and CILCOFA coefficient estimates remain statistically significant. Table 4 shows that banks using any type of interest-rate derivative, on average, experience significantly higher growth in their C&I loan portfolios. 13 This positive relationship between derivatives use and C&I loan growth is consistent with Diamond's (1984) model of financial intermediation. In that model, Diamond argues that interest-rate derivatives allow commercial banks to lessen their systematic exposure to changes in interest rates. In addition, interest-rate derivatives create extra risk tolerance, enabling banks to provide more C&I loans without increasing the total risk level faced by the banks. Furthermore, a positive and significant coefficient estimate on the DERIV variable indicates that the net effect of derivative use on C&I lending activity is complementary. That is, the complementary effect dominates any substitution effect.

Regression (2) of
The regression reported in column (3) of Table 4 examines the relative role played by each type of derivative instrument in explaining C&I loan growth. Since banks that invest in the human capital and internal control systems necessary to be active in the market for derivatives are more likely to use more than one type of derivative, 14 we also control for the effect of eleven possible interactions between each type of derivative activity in the regression. 15 The results show that the coefficient estimates on all four kinds of derivative variables are positive. The coefficient estimates on OPTIONS, FORWARDS, and FUTURES are statistically significant. These results suggest that the use of these three types (Options, Forwards, and Futures) of derivatives is significantly associated with higher C&I loan growth. 13 When the actual derivative use rather than the predicted derivative use is included in the C&I loan growth specification, the coefficient estimate (not reported) on DERIV is positive and marginally significant at the 10% level. The actual derivatives-use indicator variable is a binary variable equal one if a bank engages in any interest-rate derivative activity, or zero otherwise.
Further, except for the interaction between options and futures, none of the coefficient estimates on the interaction terms between each type of derivative activity is significant. 16 Overall, our results suggest that aggregate use of derivative instruments, in particular interest-rate options, interest-rate futures, and interest-rate forwards, is associated with higher growth rates in C&I loans.

Robustness Check
To check the validity of the regression results, we augment the regression (2) specification by adding variables measuring other characteristics of financial institutions that may explain lending activity during the sample period. The augmented regression reported in column (4) of Table 4 addresses the concern of omitting important variables that might alter the observed positive relationship between lending activity and participation in interest-rate derivatives.
First, the lagged dependent variable (LAGGED CILGA) is included in the regression to account for the possibility that the derivative-participation variable is a proxy for growth potential. we also include a control for a foreign-firm effect by introducing a binary variable equal to one if a bank is a subsidiary of a foreign financial institution (FOREIGN), or zero otherwise. Previous literature suggests that the operation of foreign-owned banks helps to fund U.S. operations of foreign industrial firms. 17 Therefore, foreign-owned banks may be expected to provide both loans and interest-rate derivatives to their customers, inducing a positive coefficient. On the other hand, foreign-owned banks also have some disadvantages due to problems in managing from a distance and coping with multiple economic/regulatory 16 Without the interaction terms, the coefficient estimates (not reported) on all four kinds of derivative instruments have a positive sign, and the estimates on options and forwards are statistically significant. 17  Finally, the ratio of the dollar value of any unused lines of credit (UNLC) to total assets is included as a measure of risk tolerance. The risk is two-dimensional. First, liquidity problems may emerge as banks commit to fill larger credit lines. Second, banks' off-balancesheet exposures to credit risk may increase as they extend lines of credit to manage the interest-rate risk. Controls introduced for these possibilities provide a way of separating loan growth from risk-taking motivations.
Regression (4) incorporates the above proxies for other activities that may cloud the positive association between derivative activity and loan growth. As shown in column (4) of Table 4, the results of the study remain robust. Specifically, the coefficient on predicted derivative activity (DERIV) remains positive and statistically significant. In addition, the coefficient on the foreign-bank variable (FOREIGN) is negative and highly significant, suggesting that foreign-owned banks experience slower growth in C&I loan lending activities.
As Brewer, Minton, and Moser (2000) suggest, dealer activities performed by the banks could give rise to a negative relationship between derivative usage and lending because 18 For example, see Berger, Dai, Ongena, and Smith (2003) and Buch (2003). 19 ISDA membership list is available at http://www.isda.org. banks enhance their revenue by acting as over-the-counter (OTC) dealers and charge a fee for placing derivative positions. Consistent with their prediction, the coefficient on the dealer variable is negative and significant. Finally, the coefficient estimate on the lagged dependent variable is not significantly different from zero. The ratio of unused lines of credit to total assets, UNLC, is positive and significant, suggesting that the higher the risk tolerance as measured by UNLC, the greater the C&I loan growth.

Conclusions
Commercial Overall, this study documents a direct relationship between derivative usage by U.S.
banks and growth in their commercial and industrial loan portfolios. More specifically, we find that aggregate use of derivative instruments, in particular interest-rate options, interestrate futures, and interest-rate forwards, is associated with higher growth rates in C&I loans.
These findings are consistent with the results of an earlier study by Brewer, Minton, and Moser (2000), who examine the relationship between lending and derivative usage for a sample of FDIC-insured commercial banks between 1985 and 1992. This documented positive association is consistent with Diamond's (1984) hypothesis that derivative contracting and lending are complementary activities. Diamond's model predicts that banks can reduce the cost of delegated monitoring by holding a diversified portfolio. Engaging in derivative activities helps banks reduce the cost of monitoring contracts issued to their loan customers, thereby enabling banks to increase their lending activities without increasing the total risk level faced by the banks.
In addition, these results suggest that C&I loan growth has a significant positive relationship with the capital ratio. These results are consistent with the previous banking research in that banks with stronger capital are more able to increase their loan portfolios. we also document a negative relationship between C&I loan charge-offs and C&I loan growth.
This negative association is in line with the notion that the charge-offs variable captures the impact of regulatory pressures or a strong economic environment, or both. Further, the main results are confirmed after a robustness check. Williams-Stanton, S., 1996, The effects of risk-based capital on wealth and risk-taking in banking, Working paper, Ohio State University.  a The average ratio of total assets equals the ratio of the notional principal amount of outstanding swaps to total assets for banks reporting the use of swaps. b The average ratio of total assets equals the ratio of the notional principal amount of outstanding options to total assets for banks reporting the use of options. c The average ratio of total assets equals the ratio of the notional principal amount of outstanding forwards to total assets for banks reporting the use of forwards. d The average ratio of total assets equals the ratio of the notional principal amount of outstanding futures to total assets for banks reporting the use of futures. a The average ratio of total assets equals the ratio of the notional principal amount of outstanding swaps to total assets for banks reporting the use of swaps. b The average ratio of total assets equals the ratio of the notional principal amount of outstanding options to total assets for banks reporting the use of options. c The average ratio of total assets equals the ratio of the notional principal amount of outstanding forwards to total assets for banks reporting the use of forwards. d The average ratio of total assets equals the ratio of the notional principal amount of outstanding futures to total assets for banks reporting the use of futures.  b CARATIO is measured as the ratio of total equity capital to total assets at time t-1. CILCOFA is measured as the ratio of C&I loan chargeoffs in period t-1 to total assets in period t-1. EMPG is the state employment growth rate, where EMP equals the total employment in the state in which the bank's headquarters are located. DERIV, SWAPS, OPTIONS, FORWARDS, and FUTURES are instrumental variables obtained from a probit specification for participation in the indicated derivative markets. OF, OS, OW, FS, FW, SW, OFS, OFW, FSW, SWO, and OFSW are eleven possible interactions between each type of derivative instruments. Among these interaction terms, O stands for option, F stands for futures, S stands for swaps, and W stands for forwards. DEALER is a binary variable, which equals to one if the institution is listed as an IDSA member, or zero otherwise. FOREIGN is a binary variable, which equals to one if the institution is a foreignowned institution, or zero otherwise. LAGGED CILGA is the first lag of the dependent variable. UNLC is unused lines of credit to total assets. The dependent variable for the first-stage regression is the probability of derivative usage. OPPRF is the sample banks' net interest margin. CARATIO is the sample banks' capital-to-asset ratio. LNTOTASST is the logarithm of the sample banks' total assets, and LAGDEPENDENT is the first lag of the dependent variable.  (2) Regression (3) Regression (4)