The risk of foreign currency contingent claims
at US commercial banks
Mukesh K. Chaudhry
a, Rohan Christie-David
b,
Timothy W. Koch
c,*, Alan K. Reichert
daNorthern State University, Aberdeen, SD, USA b
University of Southern Mississippi, Hattiesburg, MS, USA c
Darla Moore School of Business, College of Business Administration, University of South Carolina, The Francis M. Hipp Building, Columbia, SC 29208, USA
d
Cleveland State University, Cleveland, OH, USA
Received 22 December 1997; accepted 8 July 1999
Abstract
This study investigates the relationship between market-based measures of risk and foreign currency contingent claims activity at US commercial banks. Speci®cally, four types of foreign currency contingent claims are examined: purchased foreign currency option contracts, foreign-exchange swaps, commitments to purchase foreign currency and forward contracts. Within the context of the Comptroller of the Currency's (OCCÕs) Banking Circular 277, we dierentiate between the risk exposure of dealer banks and non-dealer banks. Empirical results suggest that (i) the use of options tends to increase all market-based measures of bank risk, (ii) swaps are used primarily for risk-control purposes and (iii) the use of forward contracts and currency commitments contributes mildly, if at all, to any type of risk. There is some evidence that swaps activity at dealer banks increases unsystematic risk. Otherwise, dealer and non-dealer banks appear to similarly manage foreign currency risk.Ó2000 Elsevier Science B.V. All rights reserved.
JEL classi®cation:G21
Keywords:Foreign currency; Contingent claims; Banks
Journal of Banking & Finance 24 (2000) 1399±1417
www.elsevier.com/locate/econbase
*Corresponding author. Tel.: +1-803-777-6748; fax: +1-803-777-6876. E-mail address:[email protected] (T.W. Koch).
1. Introduction
This study empirically investigates the relationship between market-based measures of risk and foreign currency contingent claims activity at US com-mercial banks. The impetus is the rapidly growing use of these o-balance sheet items and regulatory concerns regarding bank risk management practices. The oce of the Comptroller of the Currency (OCC) (1993) issued Banking Cir-cular 277 to provide guidance on risk management activities for national banks and federal branches and agencies of foreign banks entering into derivative activities. While the circular encourages banks to use derivatives for prudent purposes, it states
The complexities of ®nancial derivatives raises concerns about some insti-tutionsÕ use of derivatives under some circumstances. National banks engaging in derivatives transactions must do so in accordance with safe and sound banking practices. The OCC is concerned about how the use of ®nancial derivatives can in¯uence the risk of failure of any institution, and particularly those institutions whose failures might threaten the solvency of other institutions or negatively aect liquidity in the nationÕs ®nancial system.
More recently, central bankers from Europe, Japan and North America agreed to implement a proposal for monitoring foreign-exchange risk at banks to better assess risk exposure to interest rates and exchange rates.1 One com-ponent of foreign-exchange risk is the net open position in certain o-balance sheet instruments.
The purpose of this research is to examine the breadth of US banksÕuse of foreign-exchange contingent claims and to investigate the marketÕs perception of bank risk associated with dierent levels of usage. We speci®cally examine bank purchases of foreign currency option contracts, commitments to purchase foreign currency, and the notional principal amount of forward contracts and foreign-exchange swaps. Banks use these instruments for varied purposes, as end-users with the typical objective to reduce undesirable exposure to currency ¯uctuations, and by oering derivatives to bank customers as part of dealer operations. Banks also use these instruments speculatively to increase income. If management truly attempts to hedge, derivatives should serve to decrease bank risk, while any speculative use of these instruments will increase risk.
We extend existing research in two important ways. First, we document US banksÕuse of these four types of foreign-exchange contingent claims and, in the
spirit of Saunders et al. (1990), examine the relationship between capital market measures of bank risk and measures of this foreign-exchange o-bal-ance sheet activity for a sample of large banks from 1989 to 1993. Second, we distinguish between in¯uences across dealer banks and other banks that are active in these markets. This research is particularly important given the focus of Circular 277 on institutions whose potential insolvency might threaten li-quidity ± a clear reference to large banks that run dealer operations and/or compete across national markets. The analysis allows us to draw inferences regarding the marketÕs perception of whether banks in the aggregate use these instruments to hedge or speculate.
The relationship between risk taking by banks and their foreign currency contingent claims exposure is important to regulators and investors. Regula-tors are primarily concerned about bank safety and soundness such that the relationship between these measures and total risk will provide information about the perceived likelihood of default. The relationships between interest rate risk and foreign-exchange risk provide similar information regarding how investors react to unanticipated exposures. This is useful in assessing the likelihood that market participants will react adversely to undesired exposures which, in turn, might serve as an indicator of problem banks. While investors can diversify unsystematic risk away, it is useful to know whether any rela-tionship exists between this risk and the various exposures to identify the ap-propriate portfolio strategies. Finally, these concerns are exaggerated for dealer banks whose exposures are presumably the greatest.
Recent studies by Choi and Elyasiani (1997) and Hirtle (1997) examined similar issues with several dierences. While Choi and Elyasiani (1997) provide evidence of a link between a bankÕs derivative activity and its interest rate and exchange risk betas, we provide evidence on the association between the use of foreign currency contingent claims by dealer and non-dealer banks and mea-sures of market, systematic, and unsystematic risk in addition to interest rate and exchange risk measures. Hirtle (1997) examines the relationship between the interest-rate risk exposure of bank holding companies (BHCs) and deriv-ative usage, particularly interest rate swaps.2 We focus on foreign currency derivative activities.
2
Other studies have examined the impact of derivatives on ®rm value. Brewer and Lee (1986), Brewer et al. (1996), Carter and Sinkey (1998), Gorton and Rosen (1995), Grammatikos et al. (1986), Morgan et al. (1988), Pillo (1995), Schrand (1997) and Venkatachalam (1996) discern some relationship between the use of interest-rate derivatives and the interest sensitivity of bank or savings and loan stocks. Pillo further concludes that the stock return volatility of derivativesÕ
dealers does not increase with trading activity and suggests that dealers may better use the instruments to hedge than non-dealers. Carter and Sinkey (1998) conclude, alternatively, that the use of derivatives reduces the interest sensitivity of stock returns with similar results for dealer and non-dealer banks.
The remainder of the paper is organized as follows. In Section 2 we describe US bank activity involving foreign-exchange derivatives at year-end 1993. Section 3 presents the models used to generate the capital market measures of bank risk and relates these measures cross-sectionally to control variables and measures of bank foreign-exchange contingent claims activity. We present the empirical results in Section 4 and a summary with policy implications in Section 5.
2. US banks' use of foreign-exchange contingent claims
Table 1 presents summary data regarding the use of four types of foreign currency contingent claims at year-end 1993 for US banks. The summary statistics provided in Panel A relate to the entire banking sector, while those in Panels B and C relate to our research sample. In 1993 approximately one-third of total bank assets was controlled by banks below $10 billion in size, one-third was managed by banks with $10±50 billion in assets, and the remaining one-third was controlled by banks with assets in excess of $50 billion. It is in the two largest size groups, especially the 15 banks over $50 billion, where the vast majority of foreign-exchange contingent claim activity takes place. For ex-ample, while the 15 banks with more than $50 billion in assets manage 36.9% of total industry assets on balance sheet, they manage 76.0% of all forward-ex-change contracts, 59.7% of foreign-exforward-ex-change commitments, 86.2% of currency options and over 92% of foreign-exchange swaps.
In terms of our research sample, the 112 banks collectively represent 41.1% of total industry assets with an average bank size of approximately $15 billion as summarized in Panel B. These banks are collectively involved in 74.3% of total forward-exchange contracts, 60.6% of all foreign-exchange commitments, 82.3% of current options, and almost 95% of foreign-exchange swaps. Panel C indicates that our sample includes eleven dealer banks, all of which have assets over $50 billion. These banks account from 41% to almost 86% of the foreign currency derivatives activity.
3. Models, methodology and data
Table 1
Summary statistics for assets and foreign currency contingent claims of all bank holding companies in the US banking system 1993a(in US$ billions)
Asset range No. of
100±300 mn 690 132.66 3.2 0.005 0.04 0.001 0.01 0.091 0.01 0 0
>30 bn 383 192.37 4.7 0.055 0.47 0.587 3.91 0.210 0.02 0 0
>1±5 bn 196 445.46 10.8 0.225 1.91 0.503 3.36 2.812 0.32 0 0
>5±10 bn 47 331.10 8.0 0.075 0.64 0.180 1.20 6.037 0.70 0 0
>10±50 bn 70 1489.52 36.1 2.475 20.94 4.772 31.8 110.083 12.72 18.31 7.81
>50 bn 15 1521.72 36.9 8.985 76.01 8.946 59.68 746.538 86.23 216.08 92.15
Panel B
Banks in our sample
112 $1695.06 41.1 $8.656 74.3 $9.084 60.6 $712.31 82.3 $222.54 94.9
Panel C
Dealer banks
11 $814.86 19.8 $4.851 41.6 $6.243 41.6 $630.19 72.8 $201.35 85.9
a
This table gives aggregate dollar values (Agg) for assets and foreign currency contingent claims for the entire banking sector in 1993. The table presents the aggregate value for banks in the size category and the percentage of the total value for the entire banking sector. Curr, Comm, and Opt refer to currency, commitments, and options, respectively; mn and bn refer to million and billion, respectively.
~
Rita0ibmiR~mtbkiR~ktbeiR~et~eit; 1
where the dependent variable,R~it, is the holding period return for theith bankÕs stock in a given montht,R~mtthe holding period return on an equally weighted portfolio of common stocks,R~kta measure of unanticipated changes in the risk-free rate, R~et a measure of unanticipated changes in a weighted-average-exchange rate of the US dollar versus currencies in 10 industrial countries and ~
eitis the usual error term. In this framework,bmimeasures systematic risk and the sensitivity of the security to market-wide events, whilebki andbeimeasure the eect of unexpected changes in nominal interest rates and an index of exchange rates, respectively, on bank stock returns.
Because expected changes in interest rates and exchange rates should already be incorporated in stock returns, we construct measures of unanticipated changes for R~kt and R~et which may cause market participants to react and
potentially aect stock returns. In particular, we employ the standard ARIMA procedure to ®t a model for each of the long-term government bond rate and index of exchange rates using monthly data for 1989±1993. In both cases, an AR(1) model is appropriate.3 We then interpret the residuals from these models as the unanticipated change in the long-term rate and exchange rate, respectively, and designate the series as R~kt andR~et. Finally, we estimate the
regressions across banks as a series of seemingly unrelated regressions (SUR) to take advantage of possible contemporaneous correlation among the error terms.4
This model yields the following capital market measures of risk for each sample bank:
Dierences in the systematic risk measures across banks re¯ect dierences in the sensitivity of bank stocks to the market return, unanticipated interest rates,
rRi total return risk for banki, Var R~it
1=2
rei unsystematic risk for banki, Var ~eit
1=2
bmi systematic risk for banki
bki systematic interest rate risk for banki
bei systematic foreign-exchange risk for banki
3
We conducted the same tests throughout this research using the three-month treasury bill rate instead of the long-term government bond rate. The empirical results are virtually identical as that with the long-term rate, except an AR(3) model provided the best ®t. Thus, we do not report results using the treasury bill rate. Several researchers have used orthogonalized indices to help control for correlation among returns. We do not orthogonalize indices because Gilberto (1985), Kane and Unal (1988) and Kwan (1991) demonstrate that orthogonalization does not mitigate correlation problems and theory does not impose zero correlations.
and unanticipated-exchange rates, respectively.5 Dierences in total return and unsystematic risk, in turn, re¯ect aggregate and non-diversi®able risk. A larger (in absolute value) estimated risk measure signi®es a higher degree of bank risk along the dimension noted.
After obtaining the ®ve risk measures for each bank, we estimate three cross-sectional regression models for each risk measure. The ®rst is
Risk measurec0c1 GAP c2 CR c3 CAP c4 LIQ
c5 EFF c6 VARINC c7 SIZE c8 CI
c9 RE c10 CON c11 AGRI c12 COM
c13 FWD c14 OPT c15 SWAP ei; 2
where the risk measure is one ofrRi, bmi,bki,bei orrei. The ®rst 11 variables
control for accounting measures of bank risk from the balance sheet and in-come statement, while the last four variables represent measures of foreign currency contingent claims. The notation for the last four variables refers to a bankÕs use of commitments to purchase foreign currency (COM), currency forward contracts (FWD), purchased currency option contracts (OPT) and currency swaps (SWAP), respectively. Appendix A provides the notation and an explanation for all variables used in the regressions.
Because we are particularly interested in the relationship between the extent of a bankÕs derivativesÕusage and the marketÕs perception of risk, we construct two dierent measures of contingent claims. The ®rst is the continuous measure of a bankÕs derivativesÕactivity by type employed in the initial regression. The second groups each bank into one of three categories representing high (HIGH), medium (MED) and low (LOW) levels of contingent claim activity. To create the categorical variables, each bank in the sample is ranked ac-cording to its level of involvement relative to asset size, in each derivative activity. It is possible for a given bank to rank low in one type of foreign currency activity and high in another. Using this ratio we divide the sample into three equal parts. The top one-third of the sample includes banks where the level of derivative activity is considered to be high, the next one-third in-cludes banks considered to have a medium level of derivative activity, and the bottom one-third includes banks considered to have a low level of derivative activity. Dummy variables labeled HIGH (high), MED (medium) and LOW (low) indicate the level of activity within each type of contingent claim (e.g., HIGHCOM, MEDCOM, LOWCOM, etc.). We estimate a second set of
5Given the evidence in Kane and Unal (1988), we test for non-stationarity in the market return, interest-rate beta and exchange rate beta. The results from these tests provide no evidence of non-stationarity. These empirical results are available upon request.
cross-section regressions using these interactive measures. Finally, we estimate a third set of regressions that employs dummy variables to identify the activity of dealer banks.6We distinguish between the activity of dealer banks and non-dealer banks by employing interactive dummy variables representing non-dealer banks and the amount of their activity in each type of contingent claim.
To be included in the sample a bank holding company (BHC) had to meet the following criteria: (i) the stock was traded on the NYSE or the NASDAQ continuously during the sample period, 1989±1993 and (ii) call report data had to be available for the BHC on all variables used in the study. We exclude banks that merged or failed during the sample period.7 The data for both accounting-based and o-balance sheet variables are obtained from annual call reports and represent year-end values averaged over the ®ve year sample period (1989±1993). Stock returns are computed using monthly data obtained from the NYSE and OTC CRSP monthly tapes. Data for long-term interest rates are obtained from Ibbotson and Associates and the index of foreign-exchange value for the US dollar is from theFederal Reserve Bulletin. The returns and innovations are employed in Eq. (1) to generate estimates for each of the risk measures which serve as the dependent variables in the second stage regres-sions. The ®nal sample includes 112 BHC stocks that are actively traded on either the NYSE or the NASDAQ markets.8
An analysis of the correlation between the log of total bank assets and each independent variable reveals the importance of bank size. Speci®cally, most of the control variables exhibit a reasonably high correlation with asset size, with some of the largest correlations related to foreign currency contingent claim
6Dealer banks in the sample include: B.A. Securities, Incorporated (Bank of America); B.T. Securities Corporation (Bankers Trust); Citicorp Securities, Incorporated (Citibank); J.P. Morgan Securities, Incorporated (Morgan Guaranty Trust); NationsBanc Capital Markets, Incorporated (Nations Bank); Chase Securities, Incorporated (Chase Manhattan); C.S. First Boston Corporation (First Boston Corporation); First Chicago Capital Markets (First National Bank of Chicago); Continental Bank National Association (Continental Illinois); Chemical Securities, Incorporated (Chemical Bank); Zions First National Bank (Zions Bank).
7There is a clear supervisorship bias in this sample as banks must operate in consistent organizational form over all ®ve years to be included. The implication is that some omitted ®rms may have experienced substantially dierent risk exposures such that our estimated risk parameters do not re¯ect true exposures. The fact that we ®nd no evidence of non-stationarity inbm,bkandbe (footnote 5) suggests that our estimates are robust for the sample banks. We also conduct the same regression analysis on a year-by-year basis and discuss the results after the pooled analysis. This mitigates the impact of survivorship bias in the pooled analysis.
8
activity. For example, the correlation between the log of asset size and currency commitments is 0.496, for options is 0.467, for forward contracts is 0.449, and for currency swaps is 0.400. In addition, the correlations between bank size and the variables used to indicate the three levels of derivative involvement for each type of contingent claim also indicate a strong positive relationship as the correlation switches from being negative for the low usage group to a large positive for the high usage group. Thus, the largest banks are much more active in derivative activity than smaller institutions. These results are consistent with prior research of James (1987), Kim and Koppenhaver (1993), Gunther and Siems (1995) and Carter and Sinkey (1998) for banks and Colquitt and Hoyt (1997) for life insurance companies who ®nd a positive relationship between ®rm size and the use of derivatives.
4. Empirical results
Tables 2±6 present the parameter estimates from Eq. (2) for each of the ®ve risk measures. There are three sets of regression results in each table. The overall sample results (®rst regression) use continuous data for each of the contingent claims measures, COM, FWD, OPT and SWAP. The middle col-umns denoting the regression with the level of activity dummies reports the parameter estimates when foreign currency activity is separated into the three categories described previously. In the third regression we focus on dealer banks. Speci®cally, we use the Board of Governors of the Federal Reserve System classi®cation to identify authorized dealer banks, which we designate via a dichotomous variable (DEALER) that takes a value one for dealer banks and zero otherwise. We then investigate the impact of derivativesÕ activity at these ®rms through the use of interactive variables. The set of variables at the bottom of Tables 2±6, labeled DCOM, DFWD, DSWAP and DOPT, equals the product of the dummy variable DEALER, indicating whether or not the bank is an authorized government security dealer, and the respective contin-uous measure of each contingent claims variable.
The top portion of each table reports the results for the control variables, while the bottom portion reports results for the foreign-exchange variables. Consider Table 2 and the results of the regressions for total return risk. Here the estimated relationships among the control variables are consistent across all three models. For example, the greater is a bankÕs GAP, indicating a greater mismatch between rate-sensitive assets and rate-sensitive liabilities, the greater is total return risk. Higher relative amounts of equity capital are associated with lesser total return risk. Banks with lower personnel expense and greater earnings variability exhibit higher total return risk.
overall sample using continuous measures of contingent claims indicate that the level of foreign currency options increases total risk, while the level of foreign currency swaps lowers total risk. The levels of foreign-exchange
com-Table 2
Risk of contingent claims ± foreign currency exposure; dependent variable ± total riska
Overall sample Regression with level
CR 0.048 (0.23) 0.220 (1.05) 0.244 (1.12) CAP )1.452()5.73) )1.218()4.53) )1.195()4.45)
CON 0.035 (0.65) 0.063 (1.12) 0.010 (0.17) AGRI )0.050 ()0.39) 0.026 (0.19) )0.007 ()0.05)
In this table we report three sets of regressions. The ®rst regression is for the overall sample (identi®ed as ``overall sample''), the second accounts for dierences in levels of derivative activity (identi®ed as ``regression with level of activity dummies''), and the third takes into account the activities of dealer banks. The overall sample consists of 112 banks and includes 11 dealer banks. ***
Signi®cant at the 1% level for a two-tailedt-test. **
Signi®cant at the 5% level for a two tailedt-test. *
mitments and forward contracts, in contrast, either are not statistically sig-ni®cant or are marginally sigsig-ni®cant. Measuring contingent claim activity in categorical terms with the sample broken into thirds suggests that banks using
Table 3
Risk of contingent claims ± foreign currency exposure; dependent variable ± market riska
Overall sample Regression with level of activity dummies
Regression with dealer bank dummies
Control variables
Constant 0.644 (0.88) 1.056(1.84) 1.851(2.97)
GAP 0.011 (0.89) 0.020 (1.61) 0.013 (1.02) CR )0.639 ()0.32) 0.063 (0.03) 1.227 (0.63)
CAP )8.009()3.27) )7.091()3.12) )8.091()3.38)
LIQ )1.072 ()1.15) )1.403()1.66) )1.146 ()1.23)
EFF )0.105 ()0.71) 0.074 (0.54) )0.088 ()0.58)
VARINC 67.021(3.14) 50.592(2.52) 69.732(3.12)
SIZE 0.052(2.11)
aIn this table we report three sets of regressions. The ®rst regression is for the overall sample (identi®ed as ``overall sample''), the second accounts for dierences in levels of derivative activity (identi®ed as ``regression with level of activity dummies''), and the third takes into account the activities of dealer banks. The overall sample consists of 112 banks and includes 11 dealer banks. ***Signi®cant at the 1% level for a two-tailedt-test.
**Signi®cant at the 5% level for a two-tailedt-test. *Signi®cant at the 10% level for a two-tailedt-test.
Table 4
Risk of contingent claims ± foreign currency exposure; dependent variable ± interest rate (long-term) riska
CR 0.363 (0.12) 0.584 (0.19) 3.180 (0.92) CAP )14.011()3.81) )11.585()2.95) )8.956()2.11)
LIQ )1.129 ()0.81) )0.402 ()0.27) )0.780 ()0.47)
EFF 0.290 (1.29) 0.488(2.04)
)0.210 ()0.79)
VARINC 2.882 (0.09) 19.001 (0.55) 15.667 (0.40) SIZE )0.205(5.54) )0.018 ()0.63) )0.004 ()0.12)
In this table we report three sets of regressions. The ®rst regression is for the overall sample (identi®ed as ``overall sample''), the second accounts for dierences in levels of derivative activity (identi®ed as ``regression with level of activity dummies''), and the third takes into account the activity of dealer banks. The overall sample consists of 112 banks and includes 11 dealer banks. ***
Signi®cant at the 1% level for a two-tailedt-test. **
Signi®cant at the 5% level for a two-tailedt-test. *Signi®cant at the 10% level for a two-tailed
Table 5
Risk of contingent claims ± foreign currency exposure; dependent variable ± foreign currency riska
Overall sample Regression with level
GAP 0.032 (1.22) 0.039(1.82) 0.032 (1.18)
CR 1.504 (0.37) 1.262(1.78) 0.213 (0.05)
CAP )5.902()2.18) )5.212()1.83) )5.706()2.13)
LIQ )1.412 ()0.75) )1.317()1.69) )1.401()1.83)
EFF 0.036 (0.12) 0.239 (0.78) 0.057 (0.19) VARINC 31.295 (0.70) 39.672 (1.55) 32.502 (1.60) SIZE )0.016 ()0.45) 0.004 (0.12) )0.021 ()0.51)
CI 1.511 (1.21) 1.764 (1.48) 1.147 (0.97) RE 0.879 (0.79) 1.251 (1.15) 0.514 (0.51) CON )2.706()2.48) )2.897()2.78) )2.424()2.41)
In this table we report three sets of regressions. The ®rst regression is for the overall sample (identi®ed as ``overall sample''), the second accounts for dierences in levels of derivative activity (identi®ed as ``regression with level of activity dummies''), and the third takes into account the activities of dealer banks. The overall sample consists of 112 banks and includes 11 dealer banks. ***
Signi®cant at the 1% level for a two-tailedt-test. **
Signi®cant at the 5% level for a two-tailedt-test. *
Signi®cant at the 10% level for a two-tailedt-test.
Table 6
Risk of contingent claims ± foreign currency exposure; dependent variable ± unsystematic riska
Overall sample Regression with level of activity dummies
Regression with dealer bank dummies
Control variables
Constant 0.329(4.52) 0.083 (1.26) 0.100 (1.39)
GAP 0.003(2.27) 0.004(3.23) 0.004(3.23)
CR 0.057 (0.29) 0.186 (0.91) 0.312 (1.38) CAP )1.347()5.54) )1.094()4.19) )1.045()3.78)
LIQ )0.025 ()0.27) )0.080()0.81) )0.078 ()0.73)
EFF )0.027()1.84) )0.007 ()0.45) 0.035(1.97)
VARINC 3.378 (1.60) 2.928 (1.26) 4.266(1.65)
SIZE )0.012 ()1.53) )0.0001 ()0.05) )0.001 ()0.40)
In this table we report three sets of regressions. The ®rst regression is for the overall sample (identi®ed as ``overall sample''), the second accounts for dierences in levels of derivative activity (identi®ed as ``regression with level of activity dummies''), and the third takes into account the activities of dealer banks. The overall sample consists of 112 banks and include 11 dealer banks. ***
Signi®cant at the 1% level for a two-tailedt-test. **
Signi®cant at the 5% level for a two-tailedt-test. *
a medium level of options increase risk, whereas high levels of swaps reduce bank risk relative to the low usage banks.
The dealer bank results are similar to the results mentioned above. While options usage increases total return risk at all banks, on average, the coe-cients on DOPT indicate that currency options held by dealer banks further increase total return risk. In terms of currency swaps the results indicate that dealer bank activity signi®cantly reduces total risk relative to other banks. Forward currency contracts and currency commitments have no dierential impact at dealer versus non-dealer banks.
The results for systematic market risk, reported in Table 3, indicate that of the control variables, only equity capital and earnings volatility have a con-sistent impact on market risk. The parameter estimates for contingent claims further suggest that in the overall sample none of the four foreign currency activities are signi®cantly related to market risk. In contrast, measuring con-tingence claims activity as a categorical variable provides evidence that medi-um and high levels of options can increase market risk, whereas similar high use of commitments and swaps decreases market risk. Note that banks ranked in the middle in terms of swap use exhibit greater market risk, in contrast to the impact of higher usage. Finally, currency option activity reported by dealer banks sharply increases market risk, while swap activity has the opposite eect. Except for the control variables, the results for interest rate risk in Table 4 mimic the results for market risk. The sensitivity measure,bki, is inversely
re-lated to a bankÕs equity capital and relative sizes of the real estate and con-sumer loan portfolios. The results further indicate that none of the continuous measures of foreign currency activity impact long-term interest rate risk. When measured as categorical variables, both medium and high levels of options activity serve to increase risk whereas high levels of swaps activity appear to reduce interest rate risk. Furthermore, currency option activity reported by dealer banks increases interest rate risk, while swap activity again decreases interest rate risk relative to non-dealer banks.
Table 5 presents results related to foreign-exchange risk, which are again similar to the results for interest rate risk. Given the fact that these risks are aected by many of the same factors, the sensitivities might reasonably be related to similar ®nancial control variables and positions in foreign currency contingent claims. The sensitivity measure,bei, varies inversely with a bankÕs
and decreases with high levels of swaps. Dealer bank activity in options, in turn, is associated with an even greater level of risk while dealer swap activity is associated with lower risk.
Finally, Table 6 reports the impact of foreign currency claims on unsys-tematic risk. The results dier in several ways. First, the results for the con-tinuous claims regression indicate that both currency options and currency commitments increase unsystematic risk. Second, when the extent of currency activity is measured in categorical terms, a high level of both these instruments actually decreases unsystematic risk. Third, for dealer banks, options as well as swaps consistently increase unsystematic risk relative to non-dealer banks.
4.1. Regression analysis: Year-to-year results
The use of pooled data for 1989±1993 introduces potential survivorship bias in the sample and masks possible dierences in the relationships across years associated with structural changes in the underlying banks and economic en-vironment. To help control for survivorship bias and check robustness, we estimate the three sets of regressions in (2) on a year-to-year basis. As such, we run three regressions for each of the ®ve risk measures for each of the ®ve years, 1989±1993.9 Not surprisingly, the results in any one year dier from those reported in Tables 2±6, but support the basic conclusions.10In general, the control variables continue to aect the risk measures in a systematic way though the magnitude and signi®cance of the estimates vary. In terms of the year-to-year impacts of dierent foreign currency contingent claims, banksÕuse of currency options and swaps generally replicates the results of the pooled sample. However, commitments to purchase foreign currency are associated with higher interest rate risk at both dealer and other banks in most years, while the pooled sample suggests no relationship. Similarly, the use of currency forwards is occasionally associated with higher market risk and total risk and lower interest rate risk.
5. Summary and conclusions
This study examines the impact of four dierent types of contingent foreign currency claims, options, swaps, forwards, and commitments on various measures of capital market risk for banks. Bank risk is measured in terms of total return risk, market risk, interest rate risk, foreign currency risk, and
9The sample size varies in each year as follows: 135 (1989), 129 (1990), 125 (1991), 119 (1992) and 112 (1993).
unsystematic risk. As suggested by the 1993 Banking Circular 277, we compare the impact on a sample of dealer banks, as denoted by the Federal Reserve, with that for non-dealer banks.
Empirical results using pooled data for 1989±1993 and year-to-year data suggest several patterns in terms of market eects. In general, the use of op-tions tends to increase all types of bank risk for all banks. This should be of concern to all market participants, including bank regulators, as excessive use of these instruments can result in banks assuming high levels of risk that may not be prudent. Swaps, in contrast, seem to be used primarily for risk-control purposes. Further evidence suggests that the use of these instruments by dealer banks results in increases in unsystematic risk, which may be important to bank stock investors and to regulators who seek to maintain orderly markets. Overall, it appears that banks use currency swaps as a hedging tool while currency options are viewed as playing a more speculative role. According to tests with the pooled data, the use of forward contracts and currency com-mitments seems to contribute mildly, at most, to any type of risk. Empirical tests conducted cross-sectionally year-to-year, however, suggest that the use of these instruments does aect risk, but these impacts vary over time. The evi-dence reported in this study suggests that while all contingent claim activity does not increase risk, the OCC is correct in its concern regarding large banksÕ use of certain derivatives.
Acknowledgements
We would like to thank Leroy Brooks, Drew Winters, Kevin Jacques and seminar participants at the 1996 Financial Management Association confer-ence for helpful comments.
Appendix A. Variable de®nitions
A.Control variables
GAP net dollar value of assets less liabilities subject to repricing within one year divided by earning assets
CR loan loss provision divided by total loans (TL)
CAP book value of equity capital divided by total assets (TA) LIQ federal funds sold and short-term securities (less pledged
securities) divided by total assets
EFF salary and bene®ts divided by average assets VARINC standard deviation of net income/TA SIZE log of total assets
B.Loan diversi®cation variables
C.Foreign currency o-balance sheet variables
D. Dummy variables
References
Brewer, E., Lee, C., 1986. How the market judges bank risk? Federal Reserve Bank of Chicago Economic Perspectives 10 (6), 25±31.
Brewer, E., Jackson, W., Moser, J., 1996. Alligators in the swamp: The impact of derivatives on the ®nancial performance of depository institutions. Journal of Money, Credit, and Banking 28, 482±497.
RE real estate loans/TL
CI commercial and industrial loans/TL CON consumer loans/TL
AGRI agricultural loans/TL
COM foreign-exchange commitments to purchase divided by total assets
FWD forward contracts divided by total assets OPT foreign-exchange options divided by total assets SWAP foreign-exchange swaps divided by total assets
MEDCOM medium level currency commitment (middle one-third sample)
HIGHCOM high level of currency commitment (top one-third sample) MEDFWD medium level of forward contracts (middle one-third
sample)
HIGHFWD high level of forward contracts (top one-third sample) MEDSWAP medium level of foreign-exchange swaps (middle one-third
sample)
HIGHSWAP high level of foreign-exchange swaps (top one-third sample)
MEDOPT medium level of foreign-exchange options (middle one-third sample)
HIGHOPT high level of foreign-exchange options (top one-third sample)
Carter, D., Sinkey Jr., J., 1998. The use of interest rate derivatives by end-users: The case of large community banks. Journal of Financial Services Research 14 (1), 17±34.
Chamberlain, S., Howe, J., Popper, H., 1997. The exchange-rate exposure of US and Japanese banking institutions. Journal of Banking and Finance 21, 871±892.
Choi, J., Elyasiani, E., 1997. Derivative exposure and the interest rate and exchange rate risks of US banks. Journal of Financial Services Research 12 (2/3), 267±286.
Colquitt, L., Hoyt, R., 1997. Determinants of corporate hedging behavior: Evidence from the life insurance industry. Journal of Risk and Insurance 64 (4), 649±671.
Comptroller of the Currency, 1993. Banking Circular 277: Risk Management of Financial Derivatives.
Flannery, M., James, C., 1984a. The eects of interest changes on the common stock of ®nancial institutions. Journal of Finance 39, 1141±1153.
Flannery, M., James, C., 1984b. Market evidence on the eective maturity of bank assets and liabilities. Journal of Money, Credit and Banking 16 (4), 435±445.
Gilberto, M., 1985. Interest rate sensitivity in the common stocks of ®nancial intermediaries: A methodological note. Journal of Financial and Quantitative Analysis 20 (1), 123±126. Gorton, G., Rosen, R., 1995. Banks and derivatives. Working Paper 95-07, The Wharton School of
Business, University of Pennsylvania, Philadelphia, PA.
Grammatikos, T., Saunders, A., Swary, I., 1986. Returns and risks of US bank foreign currency activities. Journal of Finance 41 (3), 671±683.
Gunther, J., Siems, T., 1995. The likelihood and extent of bank participation in derivatives activities. Financial Industry Studies Working Paper, No. 1±95, Federal Reserve Bank of Dallas, Dallas, TX.
Hirtle, B., 1997. Derivatives, portfolio composition, and bank holding company interest rate risk exposure. Journal of Financial Services Research 12 (2/3), 243±266.
James, C., 1987. O-balance sheet banking. Federal Reserve Bank of San Francisco Economic Review 4, 21±36.
Kane, E., Unal, H., 1988. Change in market assessments of deposit institution riskiness. Journal of Financial Services Research 3, 207±230.
Kim, S.H., Koppenhaver, G.D., 1993. An empirical analysis of bank interest rate swaps. Journal of Financial Services Research 7 (1), 57±72.
Kwan, S., 1991. Re-examination of interest rate sensitivity of commercial bank stock returns using a random coecient model. Journal of Financial Services Research 5 (1), 61±76.
Morgan, G., Shome, D., Smith, S., 1988. Optimal futures positions for large banking ®rms. Journal of Finance 43 (1), 175±195.
Pillo, S., 1995. Two essays on ®nancial institutions. University of Pennsylvania, Philadelphia, PA. Saunders, A., Strock, E., Travlos, N., 1990. Ownership structure, deregulation, and bank risk
taking. Journal of Finance 45 (2), 643±654.
Schrand, C., 1997. The association between stock-price interest rate sensitivity and disclosures about derivatives instruments. Accounting Review 72, 87±109.
Tarhan, V., 1987. Unanticipated interest rates, bank stock returns, and the nominal contracting hypothesis. Journal of Banking and Finance 11 (1), 99±115.
Venkatachalam, M., 1996. Value-relevance of banks' derivatives disclosures. Journal of Accounting and Economics 22, 327±355.
