International Review of Financial Analysis 16 (2007) 1 21

Basel-2 capital adequacy: Computing the dfairT capital charge for loan commitment dtrueT credit risk
J.-P. Chateau*, J. Wu
Rouen Graduate School of Management, Mont Saint Aignan Cedex, France Received 11 August 2003; accepted 19 December 2004 Available online 8 November 2005

Abstract This research makes two contributions: (i) to price analytically put option and extension premium embedded in a borrower-extendible commitment, and (ii) to compute the dfairT capital charge that corresponds to the commitment dtrueT credit risk. In doing so, the procedure replaces the BIS accountingbased concepts of credit-conversion factor, principal-risk factor, and initial term to maturity of irrevocable commitments with the market-based concepts of exercise-cum-takedown proportion and put value implicit in the borrower-extendible commitment, respectively. Finally, the approach is developed one step further to account for the borrowers risk ratings by public credit agencies; this results in a two-dimensional (timestate of nature) risk-weighting system that applies to all commitment types. D 2005 Elsevier Inc. All rights reserved.
JEL classification: G13; G21 Keywords: Put option embedded in borrower-extendible commitments; Exercise-cum-takedown proportion; dFairT capital charge for commitment dtrueT credit risk

1. Introduction The purpose of this research is to compute the b fair Q capital charge corresponding to the b true Q credit risk of off-balance-sheet credit commitments. To do this, the creditassessment parameters mandated in Basel-11 by the Bank of International Settlement (1988, the BIS) have to be first reset in the finance well-known framework of time and states of nature.
* Corresponding author. ESC-Rouen, Bd. A. Siegfried BP 188, 76825 Mont Saint Aignan Cedex, France. Tel./fax: +33 2 32 82 57 29. E-mail addresses: jpchat@sprint.ca (J.-P. Chateau), jian.wu@groupe-esc-rouen.fr (J. Wu). 1 Although the revised guidelines (Basel-2 [2004]) are not considered per se here, our market-based concepts remain relevant to Basel-2. 1057-5219/$ - see front matter D 2005 Elsevier Inc. All rights reserved. doi:10.1016/j.irfa.2004.12.002

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

Regarding the risk dimension, commitment credit risk is measured by the credit-conversion factor (0% or 50%) and the principal-risk factor (0% or 100%). On the time dimension, commitment duration is captured by only two initial terms to maturity for irrevocable commitments those with an initial term to maturity up to 1 year or longer than 1 year. Not unexpectedly, the BIS accounting-based procedure yields a very coarse grid of credit-risk weights for off-balance-sheet loan commitments. There also are only two somewhat artificial values for both credit-conversion and principal-risk factors. With the result that the capital charge corresponding to the credit risk of longer-term irrevocable commitments is very substantial, while that for short-term irrevocable commitments, as well as all revocable commitments irrespective of their initial term to maturity, is nil. The direct linkage of commitment credit risk to its capital charge is obtained by substituting three market-based concepts to the Basel-1 accounting-based coefficients. To start with, all commitments are construed as borrower-extendible commitments and differentiated on the basis of the duration of their extension period. This replaces the distinction between revocable and irrevocable commitments, with the latter ones being further split on the basis of their initial term to maturity. Next, the credit-conversion factor makes way for an empirically relevant exercisecum-takedown proportion, namely the average amount of the credit line drawn down when the line is exercised. And finally, the principal-risk factor is replaced by the value of the put option embedded in borrower-extendible credit commitments. Valuing this implicit credit-risk derivative then raises three questions: Does the put option embedded in an extendible commitment capture its b true Q credit risk? 2) Is this value more sensitive to the time or state-ofnature parameters? And 3) how is the bfair Q capital charge for commitment b true Q credit risk actually computed? In recent years, Petersen and Rajan (1994), Ergungor (2001) or Kashyap, Rajan, and Stein (2002) among others have stressed the central role played by commitments in bank lending, and several researchers have derived alternative formulas for valuing credit-line commitments. Thakor, Hong, and Greenbaum (1981) and Ho and Saunders (1983) derived option-like values for fixed-rate straight commitments, Thakor (1982), Chateau (1990), and Chateau and Dufresne (2002) obtained put formulas for non-extendible variable-rate commitments, and Hawkins (1982) priced revolving credit lines. To the best of our knowledge, extendible or rollover commitments, namely those in which the initial commitment period is extended for another time period, have not yet been priced. Within the BIS regulatory framework, pricing borrowerextendible commitments has the advantage to circumvent the artificial dichotomy between irrevocable commitments with a 1-year initial term to maturity and those with an initial term to maturity longer than 1 year. Fortunately, there have been advances in research on derivatives with extendible maturities: Anathanarayanan and Schwartz (1980) and Athanassakos, Carayannopoulos, and Tian (1997) have priced extendible bonds, Longstaff (1990) European options with extendible maturities, and Hauser and Lauterbach (1996) extendible warrants. Here we propose a general closed-form value formula for single- or multiple-year extension premiums, of which the extendible put expression of Longstaff (1990, Eq. (12), p. 943) is but a particular case. According to Thakor et al. (1981), when the interest rate on a commitment contract is lower than that on an equivalent spot loan, the borrower receives the credit-line face value but is only indebted for its lower marked-to-market value henceforth referred to as the indebtedness value. More concretely, the borrowers claim on the lending bank constitutes an embedded, yet valuable, commitment put option. It is thus sensible to determine the impact of this implicit liability on the banks regulatory capital at the BIS audit date. The aggregate

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

face value of unused commitments is reported as an off-balance-sheet entry to the bank annual consolidated balance sheet. Yet, at the BIS annual audit date, the time remaining to commitment expiry is less than the initial term for many of the still-unused commitments. To account for this, the average time remaining to the commitment first expiry date, T 1, is standardized at T 1 s, with s = 0 denoting the date at which the BIS capital-adequacy audit takes place. Within this time frame, the commitment put option is thus European and generated by the fixed markup of credit lines with a floating prime-rate formula namely those with ba fixed markup over a stochastic index cost of fundsQ.2 In addition, its value corresponds to the banks notional liability for carrying the commitment at the audit date. Granted the above, the research sets out: (i) to value the European commitment put implicit in borrower-extendible commitments; (ii) to determine in simulation experiments the magnitude of single- or multiple-year extension premiums comprised in the extendible put values; (iii) to use the simulated values to compute the fair capital charge corresponding to the commitment true credit risk; and (iv) to propose a new two-dimensional risk-weighting system for extendible and nonextendible commitments alike. A risk-neutrality argument is used in Section 2 to value the extension premium comprised in the borrower-extendible commitment put option. A special case nested in the model is also priced, namely Thakor (1982) put option embedded in a 1-year straight commitment. A funding proportion is defined next: it combines an exercise-indicator function that captures the line exercise decision to a takedown proportion that increases with the duration of the extension period. As extendible puts are but notional values of embedded creditrisk derivatives, simulations are used in Section 3 to validate the credit-risk valuation programme. Two patterns emerge from the simulations. Put values implicit in loan commitments correspond to notional liabilities incurred by the bank for carrying unused credit lines at the BIS audit date. And these values, and hence the banks credit-risk costs, are more sensitive to risk changes in the indebtedness value than to duration variations in the extension period. Based on these simulations, the regulatory implications of extendible commitments are presented in Sections 4 and 5. To do this, the proposed procedure does away with the BIS creditconversion factor, principal-risk factor, and initial-term-to-maturity for irrevocable commitments. These parameters are replaced with the exercise-cum-takedown proportion and the put value implicit in borrower-extendible commitments; when combined together, these then directly link commitment credit risk to its market-based or dfairT capital charge. Empirically, this implies that risk-weighted balances of commitments with different duration of the extension period are positive, and so do attract a capital charge. It also ensues that (i) embedded put values constitute a finer credit-risk grid than the two artificial values of the BIS principal-risk factor, and (ii) capital charges computed from extendible-commitment balances are moderate and internally consistent for all commitment types. Section 5 finally goes one step further in accounting for the borrowers credit-risk ratings. We propose a matrix of new standard credit-risk weights sensitive to three parameters: the borrowers risk ratings of public credit agencies, the duration of the extension period, and the proportion of line funding, respectively. Interestingly, the credit-risk weights, which represent the banks notional credit-risk costs per $100 of line commitment, are more sensitive to the funding proportion that varies with the extension duration than to the
2 According to a recent Federal Reserve survey (2000), about eighty percent of U.S. commercial and industrial lending is done via loan commitments, with the vast majority being of the floating-rate type. Typically, commitments with an upfront fee only are sold to high-credit-quality firms. But longer-term commitments with upfront and rear-end fees are sold to medium size firms whose credit quality is poorer (see Berger & Udell, 1995; Petersen & Rajan, 1994). This most prevalent type is examined here.

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

borrowers credit-rating range. The computation of the credit-risk weights is summarized in a stepwise procedure. The layout of the paper is as follows. Section 2 provides the analytical value of the European put embedded in extendible and non-extendible commitments; it also determines the exercisecum-takedown proportion. Simulation results are presented in Section 3 and used in Sections 4 and 5 to quantify the link between commitment credit risk and banks capital charge. Short concluding remarks close the paper in Section 6. 2. Valuation of extendible credit commitments 2.1. The borrower-extendible commitment The salient features of a (no-default) commitment with a fixed mark-up are stylised in the decision chart below. In part (a) of the chart, the bank writes at date 0 an off-balance-sheet commitment contract for a credit line (CL) with the following features: (i) the initial 1-year commitment period, [0, T 1], is extended at T 1 for another year,3 [T 1, T 2], at the borrowers option, (ii) loan duration,4 [T 2, T *], is 1 year from date T 2 if the credit line is drawn down, (iii) the CL face value, standardized at a maximum of L = $100, remains constant over both commitment and extension periods (namely L = L 1 = L 2 = $100), and (iv) the floating primerate formula is c T 2 + m 0. Its first component, c T 2, is the banks stochastic cost of funds at exercise date T 2, with the rate on certificates of deposit (CDs) being generally used as exogenous index. The other component, the fixed forward mark-up m 0 is determined at date t = 0 when the commitment contract is written. For instance, a borrower-extendible commitment for a $100-maximum CL has a time-0 (time-T 2) floating prime rate of 6% p.a.
-----------------------------------------------------------------------------------------------------------

Decision chart of a borrower-extendible credit commitment with a fixed forward mark-up in its floating prime-rate formula. a) Initial situation at t = 0: the contractual terms of reference abstract from compensating balances, bank reserve requirement and annual fees.5

Initial 1 year 1 year extension at Duration of potential loan commitment period the borrower's option |----------------x---------------+-------------------------------+------------------------------+ s T1 T2 t=0 T* Credit line with a maximum of L = $100, fixed mark-up m0 = 1.5% p.a., and MAC clause; all three remaining in force from t = 0 to T2 f0C = 25: commitment fee fT1E = 25: extension fee fT2U = 25: non-usage fee
Later on, a multiple-year extension period is considered. But, if the commitment is exercised at date T 1, the 1-year straight commitment is a nested case examined at the end of this subsection. 4 It is explained later on in this subsection why T 2 is the exercise date here; loan duration, T 2 T*, can also be extended beyond 1 year. 5 Commitments only rarely involve compensating deposit balances (Berger & Udell, 1995), a disguised cost that is usually treated as a scaling problem. U.S. reserve requirement for transaction deposits ranges from 3% to 10% depending on the size of the financial institution (Rose, 2003). There also exist minor annual expense and annulation expense, in the order each of 35 basis points per annum (0.030.05% p.a.).
3

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

b) BIS regulatory time frame: valuation takes place at the annual audit date, s. T 1 s = 6 months is the standardized time left to the first expiry date.

The fraction 0 < i 1 of L = $100 becomes a loan ------------------------------| The borrower triggers the T* agreed upon 1-year extension or ---------------|--------------------------------| T1 T2 s=0 fT2U = 25 is paid on: (i) the unexercised lines, and (ii) the un-funded portion (1 i) of the exercised lines ---------------------------------------------------------------------------------------------------------------------fT1E = 25: extension fee
(6.5% p.a.) made up of a 4.5% p.a. (5% p.a.) stochastic cost of funds plus at both dates a fixed forward mark-up of 1.5% p.a. The fixed mark-up thus only hedges credit risk as the borrower bears the funding risk, c T 2. Most irrevocable commitments (ditto obviously for revocable ones) also comprise a material-adverse-change (MAC) clause that remains in force during both commitment and extension periods. This escape covenant allows the bank to limit or even deny credit funding if the borrowers financial condition deteriorates over both periods the degree of deterioration being judged solely by the bank from its proprietary information about the borrower.6 Thakor and Udell (1987) provide the economic rationale for the banks optimal deployment of up-front and rear-end fees in a non-extendible commitment. In their competitive equilibrium model,7 the screening device resolves the bankborrower asymmetries of information and the presence of adverse selection gives rise to split fees at the commitment end-dates. When their sorting variables are adapted to the borrower-extendible commitment, fees are deployed at the initial (t = 0), intermediate (T 1) and end (T 2) dates in part (a) of the decision chart. The first fee, C f 0 , is an upfront commitment fee of 1 / 4 of 1% per annum of the line maximum face value, namely here 25 cents per $100, the second one is an extension fee, f TE = 25 cents, of the same 1 magnitude as the initial commitment fee, and the third one is a rear-end or so-called usage fee,8 f TU. This 25-cent payment really is a non-usage fee, paid at T 2 on the un-drawn portion of the 2 credit line.
6 The borrowers ambiguously defined financial condition prevents us from defining a clear lower bound to the marked-to-market value of the credit line. Had such an explicit barrier been defined, the commitment would have automatically vanished for unsound borrowersa sort of down-and-out put option. See Lu (2004) interesting attempt at pricing commitments with a MAC covenant. 7 Borrower self-selection as a screening and risk-sharing device with optimal fee mix is also examined in Ergungor (2001), Fery, Gasbarro, Woodliff, and Zumwalt (2003), Shockley and Thakor (1997), and Thakor (1989). 8 According to Shockley and Thakor (1997, Table 1) for the years 1989 and 1990, the mean upfront fee on short-term (liquidity, working capital, and trade and finance) commitments was 24.2 basis points while the mean annual usage fee was 22.8 basis points. Regarding commitments used for general corporate purposes, the mean upfront and usage fees are 18.6 and 19.6 basis points, respectively. According to Angbazo, Pei, and Sanders (1998), both fees are declining since the mid-1990s due to strong competition; the usage fee also has a tendency to be somewhat lower than the upfront fee (Gottesman & Roberts, 2004).

Audit = valuation date

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

It now remains to explain why the put embedded in a borrower-extendible commitment is considered as a European option within the BIS regulatory time frame. The aggregate value of all unused commitments is reported as an off-balance-sheet entry to the banks annual consolidated statement. Yet, at the annual reporting date s, the time remaining to the extension date T 1 is less than the initial 1-year period for many of the outstanding rollover commitments. So, in part (b) of the chart, the average time remaining to the extension date (T 1 s) has been standardized at 6 months, with s = 0 denoting the valuation (= audit) date (see also Merton, 1977 for a similar argument). Two outcomes are possible at date T 2 for a commitment that has been extended at time T 1. It is either exercised and partial or total funding of the initial $100 results in an on-balance-sheet corporate loan it is explained in Section 2.4 how the exercise-cum-takedown proportion, 0 b p i V 1, is arrived at. Or, alternatively, the commitment simply expires and the borrower pays the non-usage fee, f TU, on the unexercised 2 line; yet he also pays this fee on the un-funded portion (1 p i ) of the exercised line. Finally, three of the decision-chart assumptions will be relaxed in the subsequent developments. There are: (i) the extension and loan-duration periods can be lengthened to 2 or more years9, (ii) the markup that captures credit risk can be adjusted by add-ons or discounts (F25 basis points, F50 basis points, and so on) for non-prime commitments,10 and (iii) higher credit spreads of nonprime commitments are associated with lower risk ratings of external credit agencies (more on this in Section 5). We now examine a particular case nested in the borrower-extendible commitment. 2.1.1. Nested case The 1-year non-extendible commitment is a special case nested in the extendible-commitment model. In that case and always within the BIS time frame, the borrower draws on the line at date T 1, with the 1-year corporate loan, [T 1, T *], becoming outstanding immediately. It ensues that the middle section in part (a) of the decision chart (relating to the extension fee and extension period) is omitted and part (b) of the chart adjusted accordingly. Thakor (1982) was the first to price the components bsplit fees + commitment putQ of non-extendible floating-rate commitments. Yet, the emergence of a 1-year straight commitment does not prevent the borrower from writing subsequently a new commitment contract. 2.2. Indebtedness value and its stochastic process Thakor et al. (1981) were the first to define the marked-to-market value of a credit line, a forward value usually referred to as the borrowers debt or indebtedness value, X. These indebtedness values, at date T 1 for a 1-year straight commitment and date T 2 for a borrowerextendible commitment, are computed, respectively, as P P 1 XT1 L1 exp m0 mT1 T1 T 4 and XT2 L2 exp m0 mT2 T2 T 4 ;

According to Fery et al. (2003, Table 1), the mean maturity of loans over the period January 1983 to December 1999 is 4.13 years for published credit lines and 4.06 years for non-published ones, respectively. This observation corroborates previous results obtained by Shockley and Thakor (1997). It thus makes sense to model 1-year commitments with extensions up to 4 or even 5 years. 10 The magnitude of such spreads over the floating prime rate is examined in Angbazo et al. (1998), Athavale and Edmister (2004), Gottesman and Roberts (2004), or Shockley and Thakor (1997).

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

Exhibit 1 Statistical analysis of X, the indebtedness-value monthly time series computed from Eq. (1) for the period 1966.01 to 2002.12 (n = 438 monthly observations) Mean Std. dev. Skewness Kurtosis 6.110 Min $96.27 Max $103.01

$99.99 0.7342 0.136 dX (t) / X (t) = 0.0000417dt + 0.0079dz (t)

Souce: Spot mark-ups, mark-up differentials, and indebtedness values computed in Eq. (1) are based on Statistics Canada monthly time series B14020 and B14043 of the prime credit rate and 90-day deposit rate of chartered (commercial) banks, respectively.

where L 1 = L 2 is the line par value, (T 1 T*) = (T 2 T*) is loan duration once the commitment has been exercised at T 1 or T 2, and (m 0 m T 1) is the difference between m 0, the fixed forward mark up set at date 0 when the commitment was written, and m T 1 = (l T 1 c T 1), the date T 1 stochastic spot mark-up defined as the difference between the spot prime credit rate, l T 1, and the funding rate in the CD market, c T 1. The same holds true for (m 0 m T 2). At date T 1 for 1-year commitments (and similarly at date T 2 for extendible ones), the commitment holder decides to draw on the line only if ceteris paribus11 m 0 b m T 1(or m 0 b m T 2), namely when the fixed mark-up is less than the stochastic spot mark-up computed from primary credit and funding rates. For instance, when our illustrative 1.5% forward mark-up is combined with, say, a 2.5% spot mark-up, the mark-up differential in Eq. (1) is negative at 1%. It follows that the inequality X T 1 b L 1 (or X T 2 b L 2) gives rise to an implicit commitment put option as the borrowers debt value is less than the option strike price. As in Thakor et al. (1981), the stochastic process of the indebtedness value is dX t =X t ldt rdzx t ; 2

where the constant terms l and r are the instantaneous drift and instantaneous standard deviation of the indebtedness-value distribution and z X (t) the Wiener process. From the statistical evidence presented in Exhibit 1, this geometric Brownian motion process is practically drift-less and so fluctuates randomly around 0%. The indebtedness value is thus the underlying claim of a forward contract as its empirical mean value, $99.99, is nearly identical to the line par value.12 So, we have to value a generalized European put option with no cost of carry, b u r d = 0, where r denotes the risk-free rate of interest and d the equivalent of a continuous dividend yield. More concretely, this describes the situation in which the bank does not hedge fixed mark-up and commitment funding at the underwriting date. At the actual funding date, the bank uses its own available demand deposits and/or sells CDs in the spot (wholesale) market. The valuation procedure below takes these bank-specific features into account. 2.3. Valuing the European put embedded in extendible and non-extendible commitments We now price in turn the European put option implicit in 1-year straight commitments and commitments with single- or multiple-year extension periods.

11 The alternative approach is the all-in-cost basis, in which markup fees are computed and compared for credit commitments and spot loans. 12 Although the indebtedness value is not likely to trade directly, the difficulty is overcome by observing that the spot mark-up is a bquasi-priceQ as it results from the actual (equilibrium) prices in the banks primary lending and funding markets. Consult Chateau and Dufresne (1998) for its detailed computation.

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

2.3.1. Pricing the put embedded in a 1-year straight commitment The date-0 value of the put option embedded in a 1-year non-extendible commitment is labelled P gbs, as it is isomorphic to the value of the generalized BlackScholes European put option with cost of carry b = 0, as d = r. So13: Pgbs X ; L1 ; T1 erT1 E maxfL1 XT1 ; 0g erT1 L1 N d X N d ; 3

where dF = {ln (X / L 1) F .5r 2T 1} / (rMT 1). In Eq. (3), E denotes expectations in the risk-neutral world and N (dF) is the cumulative probability at dF of the standard normal density, the other terms having been defined previously. The value of the implicit commitment put thus depends on the date-0 indebtedness value X, the credit-line par (= exercise) value L 1, and the maturity date T 1. 2.3.2. Pricing the borrower-extendible commitment put We denote the value of the European commitment put embedded in borrower-extendible commitments as EPi(X T 1, L 1, T 1, L 2, T 2, f TE); it depends on the indebtedness value at date T 1, X T 1, 1 the credit-line exercise value L 1 = L 2, the maturity dates T 1 and T 2, and the date-T 1 extension fee, f TE. Postscript i to EP refers to the duration of the extension period, T 2 T 1, which varies with 1 T 2 (from 2 to 6 years) as T 1 is fixed at 1 year. For instance, EP1(u) refers to the 1-year extension period of a 2-year commitment contract. The value of the extension privilege is thus defined as:   EPi-premium EPi XT1 ; L1 ; T1 ; L2 ; T2 ; fTE Pgbs X ; L1 ; T1 ; 1 where the right-hand-side second term is priced in Eq. (3). We now call upon a risk-neutral argument to value the extension privilege. This premium is the expected value of the extension privilege at time T 1 in a risk-neutral world, discounted to the valuation time s = 0 at the risk-free rate of interest, r. Namely EPi-premium erT1 E EPi-premium at T1 n o! rT1 E e E max 0; Pgbs;T1 XT1 ; L2 ; T2 T1 maxfL1 XT1 ; 0g fT1 ; 4 value of a generalized BlackScholes put option at date T 1. The where Pgbs;T1 (u) is the n h oi extension payoff is max 0; Pgbs;T1 XTL2 ; T2 T1 L1 XT1 fTE ; when X T 1 V L, and 1 1 max 0; Pgbs;T1 XT1 ; L2 ; T2 T1 fTE when X T 1 N L. 1 As our representative bank charges a positive extension fee, f TE = 25 cents, the commitment 1 contract is extended at T 1 only if the indebtedness value lies in the extension interval, [I 2, I 1], with I 2 b I 1. This requires two sets of conditions: (i) a maturity condition regarding the extension region, and (ii) an extension-fee condition governing the bounds as well as the magnitude of the extension interval. In the first case, if X T 1 b I 2, the commitment put is exercised rather than extended at time T 1, but if X T 1 N I 1, the put is allowed to expire out of the money at time T 1.

Commitment put options are valued at time zero unless stated otherwise explicitly; the time subscript to X is dropped when there is no ambiguity.

13

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

Clearly, the extension privilege is valuable only when I 2 b X T 1 b I 1. The interval upper and lower bounds, I 1 and I 2, can be found by solving the following two equations at date T 1: Pgbs I1 ; L2 ; T2 T1 fTE 0 1 and Pgbs I2 ; L2 ; T2 T1 fTE L1 I2 0: 1 In the second case, the following sufficient condition must hold for the extension fee: If fTE bPgbs L1 ; L2 ; T2 T1 ; 1 then I2 bI1 : 7 6 5

More concretely, the length of the [I 2 b I 1] region increases with the duration of the extension period, T 2T 1, in the generalized BlackScholes put value. The condition I 2 b I 1 is equivalent to the following inequalities: I 2 b L 1 and L 1 b I 1. By contrast, when the fee inequality is reversed in Eq. (7), the extension region becomes [I 2 N I 1] and the value of the extension premium turns negativea generally unattractive situation. Once these conditions are satisfied, the final payoff of the extension privilege is written: h i Pgbs;T1 XT1 ; L2 ; T2 T1 fTE L1 XT1 1I2 V XT1 V L1 1 h i Pgbs;T1 XT1 ; L2 ; T2 T1 fTE 1L1 V XT1 V I1 ; 8 1 where 1condition is equal to one if the condition is verified and zero otherwise. Following some tedious but straightforward developments reported in Wu (1997), the closed-form expression for the value of the extension premium is14: p p EPi-premium X edT2 N2 x4; z1 ; q L2 erT2 N2 x4 r T2 ; z1 r T1 ; q p p X edT2 N2 x4; z2 ; q L2 erT2 N2 x4 r T2 ; z2 r T1 ; q p fTE erT1 N z1 rT1 fTE erT1 N z2 r T1 Pgbs X ; L1 ; T1 1 1 p 9 X edT1 N Z2 L1 erT1 N z2 r T1 ; where the terms l, q, x, x*, z 1 and z 2 are defined as follows: l x z1 q x* z2 r d .5r 2 [ln (X / L 1) + (l + r 2)T 1] / [rMT 1] [ln (X / I 1) + (l + r 2)T 1] / [rMT 1] (T 1 / T 2)1 / 2 [ln (X / L 2) + (l + r 2)T 2] / [rMT 2] [ln (X / I 2) + (l + r 2)T 1] / [rMT 1]

In Eq. (9), N[u] is the cumulative probability of the standard normal density and N 2(u, u, q) is the cumulative probability of the standard bivariate normal density with correlation q.
14 Longstaffs extendible put option (1990, Eq. (12), p. 943) obtains when simultaneously d = 0 and the standard Black and Scholes (1973) put formula is used for P gbs(u) in Eq. (9). But, even in that case, it may not satisfy the extension-fee sufficient condition: the inequality in Eq. (7) only holds true for very small extension fees, namely here smaller than the 25-cent extension fee representative of credit commitments. For increasing fees, the extension region first collapses to a point before reversing to [I 2 N I 1].

10

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

Clearly, adding the value of the extension premium, EPi-premium, to that of the 1-year straight commitment put, P gbs, yields the extendible put value. We so verify that   10 EPi XT1 ; L1 ; T1 ; L2 ; T2 ; fTE Pgbs X ; L1 ; T1 EPi-premium with i: 1; . . . ; 5 1 or the borrower-extendible put value = the 1-year put value + the value of the single- or multipleyear extension premium. The very general equation (9) can also be used to price extendible put options on nondividend-paying stocks, stocks paying a continuous dividend yield, forward and futures contracts, and currency contracts. Each valuation depends simultaneously on the value of the cost-of-carry term, b, and the single-period European put value. For instance, when b = r and P gbs = P BS, Eq. (9) determines the value of an extendible put option on a non-dividend-paying stock. When b = 0 and the one-period put is estimated with the Black (1976) model, Eq. (9) gives the value of a put option on a forward or futures contract. When b = r d and the one-period put is computed with the Merton (1973) model, Eq. (9) gives the value of a stock option with continuous dividend yield. Finally, when b = r r f and the one-period put is estimated with the Garman and Kohlhagen (1983) model, Eq. (9) gives the value of a currency put option. 2.4. Modelling the line-funding proportion Once the embedded put value is computed, it remains to determine the proportion of the credit line taken down. Recall how this was formalized in part (b) of the decision chart in Section 2.1: first the commitment is exercised or not, and next for those exercised, partial or total funding occurs. Modelling the exercise-cum-funding proportion of individual commitments in the BIS regulatory framework is exceedingly difficult. Some of the reasons of the difficulty are: (i) commitments have different initial starting dates and maximum amounts, (ii) draw downs take place on different dates, (iii) some lines are completely drawn down, others are partially drawn down and in stages, and some are left unexercised altogether, and (iv) banks take advantage of the MAC clause to limit or even cancel funding. So, we propose the following bank-level solution: (i) the $100 commitment functions as a standard unit that is either completely drawn down or completely left un-drawn,15 and (ii) the exercise-cum-takedown proportion applies at the bank level to the aggregate amount of commitments within each class or extension period, namely to the dollar total of each class of commitment at the audit date. The conditional average exercise-cum-takedown proportion p i is formalized as follows: p E di 1Xi V L di P Xi V L; 11

where d i is the funding proportion, 1condition is equal to one if the condition is verified and zero otherwise, and i: 0, . . .. , 5 refers to the duration of the extension periodfrom 0 year for straight commitments to 5 years for 6-year commitments with a 5-year extension period. The complementary proportion is thus 1 pi 1 di P Xi V L P Xi NL: 12

More concretely, the total amount of commitment funding in class i is computed as the average proportion p i times the aggregate amount of commitments within the extension period under
15

For the sake of simplicity, we have reallocated partial takedown to these two proportions.

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

11

consideration; and (1 p i ) applied to the same-class aggregate amount determines its total unfunded amount (from totally un-funded lines as well as from the un-drawn portions of the exercised commitments). From the empirical evidence reported in Morgan (1993)16 and later on in panel B of Table 2 for a large international bank, we select a proportion p i , that remains constant within each extension period but increases with the duration of the extension period. To wit, a proportion of p 0 = 50% means that 50% of the aggregate contractual value of all commitments with an initial maturity less than 1 year is taken down; this proportion increases to p 5 = 75% for commitments with a 5-year extension period. The percentage chosen for straight commitments or those with short extension periods is relatively low because: (i) there is little time left to draw down these credit lines, and (ii) the borrower intentionally refrains from taking down full funding so as to avoid being charged higher commitment fees in the next period (Ergungor, 2001). The funding proportion is likely to increase, however, as borrowers have more time and investment opportunities to draw down (even cumulatively) longer-term irrevocable credit lines. The exercise-cum-takedown proportion just defined along with the commitment put value is all that we need to compute the banks capital charge corresponding to the commitment credit risk. Eqs. (1)(12) form the credit-risk valuation programme of off-balance-sheet commitments, which is estimated in the next section. 3. Simulation results 3.1. Simulation As embedded credit-risk derivatives, the put values implicit in extendible and nonextendible commitments alike are but notional values. We thus rely on simulations to compute their values, and our simulation parameters are based on the statistical evidence reported in Exhibit 1 in Section 2.2. The latter shows that historically the indebtedness value X varies in the value range $96.3 to $103: it is thus sensible to set X at $100, $99.5, $99, $98.5 and $98, for a commitment put that moves progressively in the money.17 With a line par value of $100, the slightly in-the-money indebtedness values simulate small increases in the spot markup of the class of floating prime-rate borrowers over the commitment and/or extension periods. Granted these indebtedness values, simulation experiments are performed for 1-year nonextendible commitments, P gbs, and borrower-extendible commitments, EPis, with extension period from 1 to 5 years. The time to commitment maturity at valuation date s = 0 is T 1 s = 0.5 year for the 1-year straight commitment and T 2 s = 1.5, ..., 5.5 years for the extendible commitments contracts. In addition, the following parameters are common to all simulations: the credit-line strike price remains constant through time, L 1 = L 2 = $100, the riskfree interest rate, r = 0.04, is consistent with the 4.5% CD rate introduced in Section 2.1, and q = (T 1 / T 2)1 / 2. The indebtedness-value volatility reported in Exhibit 1 is r = 0.7342 or 2.54%

Morgan (1993) indicates that between 1988 and 1990, the fraction of the loan limit actually borrowed by prime-rate borrowers is about 55%; unfortunately, he is not reporting the number of commitments left unexercised. In Athavale and Edmister (2004), the dollar amounts of loans are reported, but not the commitment initial size; conversely the facility sizes are reported in Gottesman and Roberts (2004), but not the funding amounts. 17 To consider values below $98 is of limited interest since there are only three values lower than $98 out of 438 observations.

16

12

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

on an annual basis; the simulations are performed with a 3% annualised volatility, namely r = 0.03. Before reporting on the simulations, we first clarify the meaning of computed put values. Consider the very plausible scenario represented by the entries in column (3) of Table 1, in which the indebtedness value is slightly in the money at X = $99. According to the first boldfaced entry in column (3), the estimate P gbs = 1.406 means that the European put embedded in a 1-year straight commitment has an equilibrium value of 1.41% of the $100 par value if: (i) the floating prime-rate commitment with a 1.5% p.a. fixed forward mark-up is priced when the stochastic mark-up on spot loans is 2.5% p.a., and (ii) the remaining life of contract is 6 months. By way of contrast, when the time remaining to commitment expiry is 18 months, the put value of the borrower-extendible commitments in row (1a) of matrix 1 is $1.722. Put values of longer-term commitments are also computed, with $2.45 corresponding to a commitment with a 5-year extension period. The magnitude of the extension premiums comprised in borrower-extendible commitments is shown in matrix 2. For our reference scenario in column (3) again, the value of the extension privilege is a modest $0.32 (or 18.3% of the extendible put value) when the duration of the extension period is 1 year, but much steeper (up to $1.03 or 42.5%) for longerterm extendible commitments.

Table 1 European put values embedded in extendible and non-extendible commitments (1) X P gbs Matrix 1 a) EP1 b) EP2 c) EP3 d) EP4 e) EP5 Matrix 2 a) EP1-Premium in % b) EP2-Premium in % c) EP3-Premium in % d) EP4-Premium in % e) EP5-Premium in % in $ 98.0 2.148 2.353 2.608 2.733 2.859 2.953 0.205 8.7 0.460 17.6 0.585 21.4 0.711 24.9 0.805 27.3 (2) 98.5 1.759 2.007 2.268 2.450 2.578 2.687 0.248 12.4 0.509 22.4 0.691 28.2 0.819 31.8 0.928 34.5 (3) 99.0 1.406 1.722 1.990 2.185 2.354 2.446 0.316 18.3 0.584 29.3 0.778 35.6 0.947 40.2 1.034 42.5 (4) 99.5 1.095 1.435 1.715 1.937 2.115 2.209 0.339 23.7 0.619 36.1 0.841 43.1 1.019 48.2 1.114 50.4 (5) 100 0.830 1.143 1.496 1.717 1.894 2.006 0.314 27.4 0.667 44.6 0.888 51.7 1.064 56.2 1.176 58.6

Valuation at date s = 0. Entries on line 1: P gbs, European put values implicit in 1-year straight commitments, from Eq. (3). Entries in matrix 1: put values implicit in borrower-extendible commitments, EPi, from Eq. (10) with i: 1,. . ., 5 the length of the extension period. Entries in matrix 2: extension premiums (from Eq. (9)) and as a percentage of the extendible put value, [EPi P gbs] / EPi, with i: 1, ., 5, respectively. Parameter definition: d=equivalent to a continuous dividend yield, in % per annum; L 1 = L 2: credit line exercise value in $; r=short-term rate of interest, in % per annum; r = indebtednessvalue volatility in % per annum; T 1=commitment initial maturity date; T 2=commitment terminal maturity date; T 2 T 1=duration of the extension period, from 1 to 5 years; T*=loan maturity date; and X=indebtedness value in $ computed from Eq. (1). Common parameter values: L 1 = L 2 = 100; d = r = 0.04; r = 0.03; T 1 s = 0.5; T 2 s = 1.5, . . ., 5.5; T 2 T 1 = 1, . . ., 5.

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

13

3.2. Credit-risk assessment in terms of commitment put values and extension premiums Two revealing patterns of credit-risk assessment are emerging from the matrices of Table 1. The first tendency focuses on the notional liability incurred by the bank for carrying unused credit lines at the audit date. Visual inspection of matrix 1 shows that commitment put values are more sensitive to risk variations (moving from right to left in the rows) than to time variations (going down each column). To wit, in row (1a) for a commitment that offers a 1-year extension, the put value increases steadily from approximately $1.14 for X at the money to about $2.35 for an in-the-money indebtedness value. The other rows depict similar put-like value curves, which are shown graphically in Fig. 1. By way of contrast, the matrix columns capture the effect of the extension duration on the put values; for instance, for X = $98 in the matrix first column, put values are slowly increasing from $2.35 for a 1-year extendible commitment to $2.95 for a commitment with a 5-year extension period. This also corresponds in Fig. 1, for a given indebtedness value, to a move from one value on a lower put-like curve to another one on a higher put-like curve. In brief, matrix 1 of Table 1 and Fig. 1 clearly indicate that put values, and hence bank credit-risk costs, are more sensitive to risk changes in the indebtedness-value than to duration variations of the extension period. The other pattern, revealed from the rows and columns of matrix 2, is that the extension premiums expressed as a percentage of the EPi-values are: (i) increasing with the duration of the extension period, but (ii) declining when the indebtedness value moves deeper in the money. According to entries on row (2a) for instance, the 1-year extension premium as a percentage of EP1 declines from 27.4% to 8.7% when the indebtedness values move deeper in the money. This declining pattern is duplicated in each of the other rows of matrix 2, but from a higher starting point. In other terms, the extension premiums implicit in commitments with longer extension periods are percentage-wise much larger than those embedded in commitment with short extension periods; yet the declining pattern becomes smoother for longer extension periods. These simulation results are used in the next two sections to quantify the link between commitment credit risk and its risk-weighted capital charge.

3.5 $-values of Pgbs and EP1 to EP5 3 2.5 2 1.5 1 0.5 0

2.953 2.859 2.733 2.608 2.353 2.148 2.007 1.759

2.687 2.578 2.45 2.268

2.446 2.354 2.185 1.99 1.722 1.406

2.209 2.115 1.937 1.715 1.435 1.095

2.006 1.894 1.717 1.496 1.143 0.83

Pgbs EP1 EP2 EP3 EP4 EP5

98

98.5

99

99.5

100

X = indebtedness values: from $100 (par value) to $98 (slightly in the money).

Fig. 1. The put-like curves depicted by straight, P gbs, and extendible, EP1 to EP5, put values (in $). The extension premuim is the vertical value difference of each EPi with P gbs: to wit at X = $100, EP1 P gbs = $1.143 $0.83 = $0.313.

14

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

4. The dfairT capital charge for commitment dtrueT credit risk Basel-1 credit risk rules (BIS, 1988) require that standard risk-adjusted balances be determined for each off- as well as on-balance-sheet instrument and their aggregate value be weighted against a definition of regulatory capital. To calculate risk-adjusted values, offbalance-sheet contractual amounts are initially converted by way of credit-conversion factors to on-balance-sheet bcredit-equivalent amountsQ; which in turn are weighted by appropriate principal-risk factors to determine brisk-adjusted balancesQ. Since the end of 1992, a minimum total capital requirement of 8% applies to such balances. Revised guidelines were adopted in 2004 (BIS, 2004). With respect to Basel-1, the revised standardized approach introduces two significant changes for off-balance-sheet commitments. There are now three credit-conversion factors (0%, 20% and 50%) and the principal-risk factor for short-term corporate commitments is now set at 100%, but can be reduced to 75% for the banks retail portfolio under certain regulatory and granularity conditions (BIS, 2004, pp. 230233). In this section and the next one, the Basel-2 revised guidelines are not taken into account, as they will no be in force before late 2006.18 Yet our policy implications remain pertinent for Basle-2. The information regarding commitments is presented in Panel A of Table 2 for a large international bank, the Royal Bank of Canada, as at October 31, 2003. Consider first how irrevocable commitments with a 1-year initial term to maturity and all revocable commitments without distinction of initial term to maturity are treated under the BIS present guideline. According to line (3) of Panel A, the commitment credit-equivalent amount of both types is nil, since the credit-conversion factor applied to their contractual amount ($40.4 billion and $58.9 billion, respectively, on line (1)) is 0%; and their risk-adjusted balance on line (5) is accordingly also nil as their risk factor is also 0%. The resultant regulatory capital charge (line (5) 8%) is thus also nil. There is thus no link between actual and/or potential credit risk and the bankTs capital charge: short-term irrevocable commitments and all revocable commitments do not affect the risk-adjusted capital requirement at any point in time or on a continuous basis. The situation is not the same however for longer-term irrevocable commitments and on-balance-sheet loans. On line (5), the risk-adjusted balance of over-1-year irrevocable commitments is $13.4 billion and that of other (mainly corporate) loans is $82.2 billion, and in both cases, the theoretical principalrisk weight is 100% according to line (4). On line (1) also, the aggregate of all commitment contractual amounts, $128.4 billion, is slightly larger than the total amount of on-balancesheet corporate loans, $113.7 billion shown on line (3). But $100.2 billion or 78% of all offbalance-sheet commitments is deemed risk-less according to the BIS accounting-based valuation of commitment credit risk. Thus, this nil risk-weighted balance along with the extremely large risk-adjusted balance ($13.4 billion) for longer-term irrevocable commitments defines but a very coarse credit-risk grid. Panel B of Table 2 complements the previous panel. According to this panel, revocable and irrevocable commitments are classified in four time ranges defined on the basis of their initial term to maturity. These ranges are particularly relevant since we intend to differentiate borrower-

18 An analysis and critique of the BIS revised guidelines is found, among others, in Andre, Matthieu, and Zhang (2001), Barrios and Blanco (2003), Decamps, Rochet, and Roger (2004), Ferguson (2003), Fisher (2001), Hammes and Shapiro (2001), Krahnen and Weber (2001) or Santos (2001).

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

15

Table 2 BIS current accounting-based computation of risk-weighted balances of off-balance-sheet commitments and on-balancesheet loans Panel A Off-balance-sheet commitments Irrevocablea With an original term to maturity (1) Contractual amount, C$ in billions (2) Credit conversion factor, in % (3) Credit-equivalent amount, C$ in billions (4) Principal risk factor, in % (5) BIS risk-adjusted balance, C$ in billions Panel B BIS time ranges Irrevocable commitments C$ in billions Revocable commitmentsa C$ in billions Sub-total, C$ in billions Each sub-total as a % of total
a a

Revocablea N1 year 28.2 50% 14.1 100 / 95%c 13.4 59.8 0 nil 0 nil

On-balance-sheet loans n.ab n.ab 113.7 100 / 72%c 82.2

b1 year 40.4 0 nil 0 nil

within 1 year 40.3 59.8 100.1 78.0

1 to 3 years 19.1 19.1 14.9

over 3 to 5 years 5.2 5.2 4.0

over 5 years 4.0 4.0 3.1

Total 68.6 59.8 128.4 100.0

Irrevocable commitments are unused portions of firm authorizations to extend credit and revocable commitments are offers but not obligations to extend credit. b n.a.=not applicable. c The first figure refers to the BIS-set percentage and the second to the actual (after netting out) weighted average of counterparty risk within this class. The latter figure is used to compute line (5) in Panel A. Source: Royal Bank of Canada, 2003 annual report, Canadian GAAP. For Panel A: Table 21, p 60A and Note 20, p 94A; and for Panel B: Table 24, p 65A.

extendible commitments on the basis of their extension duration. The panel inspection reveals that most of the commitments are irrevocable ones, except for very short-term ones, of which the majority is of the revocable type. In aggregate, 92.9% of all commitments have an initial maturity up to 3 years, and only 3.1% of them have an initial term to maturity longer than 5 years. We are now in a position to offer an alternative to the BIS assessment of commitment credit risk. It is based on three premises. Firstly, the use of borrower-extendible commitments allows us to circumvent the BIS dichotomy of revocable and irrevocable commitments, the latter being further split on the basis of only two initial-term-to-maturity. Secondly, the proportion of [off-balance-sheet] commitments that is likely to become [onbalance-sheet] outstanding loans is captured by the exercise-cum-takedown proportion, which depends on the commitment maturity ranges introduced in Panel B of Table 2. And thirdly, the commitment credit risk is determined by the put value implicit in borrowerextendible commitments. In other terms, the exercise-cum-takedown proportion and the extendible-commitment put value play the role of the BIS credit-conversion factor and principal-risk factor, respectively. The approach is illustrated in Table 3, where the benchmark scenario X = $99 from Table 1 is combined with data from Table 2. As the computation is for illustrative purpose only, we select the mid-point of the time ranges from Panel B of Table 2; for instance all 1-to-3-year commitments (Panel B subtotal, L i = $19.1

16

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

Table 3 Fair (option-based) capital charge of off-balance-sheet commitments; all commitments are assumed to be borrowerextendible commitments Original term to maturitya (1) (2) (3) (4) (5) (6) Contractual amount, L i in C$ in billions Exercise-cum-takedown proportion, p i in % Credit-equivalent amount, C$ in billions Commitment put value, EPi per C$ billion b Risk-weighted balance, C$ in millions Fair capital charge, C$ in millions 1 year 100.1 50 50 .01416 703.7 56.3 2 years 19.1 55 10.5 .01722 181.0 14.5 4 years 5.2 65 3.4 .02185 73.9 5.9 6 years 4.0 75 3.0 .02446 73.4 5.9

Or analytically: L i p i EPi 0.08 = the bfairQ capital charge, expressed in dollars, corresponding to the btrueQ credit risk of commitments in the ith maturity range. a The original term to maturity corresponds to the mid-point of the maturity ranges introduced in Panel B of Table 2. To wit, the 2-year commitment (1-year commitment period + 1-year extension period) characterizes commitments with an initial term from 1 to 3 years. b These EPi values are from column (3) in Table 1, but per one billion of commitment face value. They correspond to our representative scenario in which the indebtedness value is slightly in the money at X = $99.

billion) are considered as commitments with an average 2-year maturity. The computation is as follows: 19:1 billion 0:55 10:51 billion $10:51 billion 0:01722 commitment put value per $ billion $181 million $181 million 0:08 $14:52 billion: On the first line, the exercise-cum-takedown proportion of 55% converts the offbalance-sheet contractual amount into an on-balance-sheet credit-equivalent amount-also reported on line (3) of Table 3. The latter result is then multiplied by the European put value (the credit risk embedded in 1-to-3-year extendible commitments) to arrive on the second line at the risk-adjusted balance of short-term commitments an amount also shown on line (5) of Table 3. On the third line finally, the capital charge obtains by applying the 8% capital requirement to the just-computed risk-weighted balance this corresponds to line (6) in Table 3. In Table 3 similarly, the fair capital charge is $56.3 million for straight commitments, $5.9 million for all commitments in the 3-to-5-year maturity range, and another 5.9 million for all commitments with maturities over 5 years. It thus appears that the capital charge for commitments with different original term to maturity is usually moderate and internally consistent all in the order of a couple of millions. This is to be contrasted with the BIS dichotomy of no capital charge for both non-extendible irrevocable commitments and all revocable commitments, and an extremely substantial one (0.08 $13.4 billion = $1.07 billion) for longer-term irrevocable commitments. This alternative procedure can still be developed one step further. 5. New standard credit risk weights for loan commitments We now propose that the risk weights applicable to all credit commitments be based on EPi = f (X, r, r, L 1, L 2, T 2 T 1, f TE) and p i = g (d i , 1condition, T 2 T 1). In the previous develop1 ments, most of the parameters of both variables were kept constant, with only X and T 2 T 1

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

17

being really variable. Ceteris paribus we can thus determine the sensitivity of EPi with respect to X and T 2 T 1, and that of p i in terms of T 2 T 1. And the resultant risk-weighting system relies on two not-unreasonable assumptions: (i) the extendible put value is mainly a function of the indebtedness value, and (ii) the funding proportion varies with the length of the extension period. Firstly, regarding the extendible-put sensitivity to X, we make the following observation: the floating credit rate and hence forward mark-up of line commitments are generally set below the credit rate and mark-up set in spot loans. It is moreover sensible to assume that the differential between spot mark-up and forward mark-up grows larger as the borrowers risk rating by external credit agencies declines. In essence, we propose to associate the progressively in-the-money indebtedness values with the declining risk-rating ranges proposed for on-balance-sheet loans in the Second Consultative Document (Bank for International Settlements, 2000 or Fischer, 2001). The argument runs as follows. For prime-rate borrowers (say, those in the credit-risk range [AAA to AA-]), the bank is likely to charge a spot mark-up that is equal to the forward mark-up of a credit commitment: in that case the indebtedness value, X, is equal to the line par value, L. But for spot loans and credit lines of borrowers with a risk rating in the range or risk bucket [A+ to A], the loan spot mark-up is slightly higher than the corresponding forward mark-up charged on credit lines. Hence according to expression (1) the indebtedness value corresponding to this risk bucket is lower than the line $100 par value, say $99.5. To lower risk ranges correspond progressively deeper in-the-money indebtedness values. This holds true up to the lowest risk range, defined as less than B-, which corresponds to the indebtedness value $98. Secondly, as it was already mentioned in Section 2.4, the proportion of line funding is likely to be somewhat greater the longer the commitment extension period. Generally speaking, borrowers have more opportunities to draw cumulatively on the credit line if the extension period is longer than 1 year. The computation of the new risk weights is based on the following scale: the funding proportion increases progressively from 50% of the initial $100 maximum for 1-year straight commitments to 75% for 6-year commitmentsnamely those with a 5-year extension period beyond the initial 1-year commitment. Given the above assumptions, the proposed matrix of standard risk weights has the advantage of being a function of three parameters: the risk rating ranges of external credit agencies, the duration of the extension period, and the variable takedown proportion. The granularity of this two-dimensional system is indeed richer (although it could be improved by increasing the number of risk grades in a banks

Table 4 Two-dimensional risk-weighting system: new standard credit-risk weights per $100 of straight and borrower-extendible credit commitments (up to a 5-year extension) Borrowers risk bucket or Indebtedness value X, in $ Line takedowna, p i in %: EP5; p 5 = 0.75 EP4; p 4 = 0.70 EP3; p 3 = 0.65 EP2; p 2 = 0.60 EP1; p 1 = 0.55 P gbs; p 0 = 0.50 bB $98.0 $2.215 1.858 1.776 1.565 1.294 1.074 [BB+ to B] $98.5 2.015 1.805 1.592 1.361 1.104 0.879 [3B+ to 3B] $99.0 1.834 1.648 1.420 1.194 0.947 0.703 [A+ to A] $99.5 1.657 1.481 1.259 1.029 0.789 0.547 [AAA to AA] $100.0 1.504 1.326 1.116 0.898 0.629 0.415

Common parameter values: L 1 = L 2 = 100; d = r =0.04; r = 0.03; T 1 s = 0.5; T 2 s = 1.5, . . ., 4.5. a The takedown proportion varies with the duration of the extension period, denoted by the p-subscript.

18

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

2.5

2 Bank's credit-risk cost per $100 of line commitment 1.5 2-2.5 1.5-2 1-1.5 0.5-1 0-0.5

0.5

0 BBB+; B3B+; 3BA+; A-

3As; 2A-

EP5;0.75 EP4;0.70 EP3;0.65 Y base axis: takedown % which EP2;0.60 varies with the extension EP1;0.55 duration; from Pgbs;0.50 to EP5;0.75 Pgbs;0.50

X base axis: Borrowers' risk buckets from [3As to 2As-] for X = $100 to less than [B-] for X = $98

Fig. 2. Sensitivity of credit-risk weights to: (i) the borrowers risk buckets (on the X base axis) and (ii) the takedown proportion that varies with the extension duration (on the Y base axis).

internal-rating system) than the present BIS coefficients characterized by the superficial time-tomaturity dichotomy for irrevocable commitments and their only two principal-risk factors. The new risk weights per $100 of borrowers extendible commitment are presented in Table 4: columns of this time-risk table refer to rating ranges from public credit agencies and rows to the funding proportion that varies with the extension duration.19 Not unexpectedly, the table rows reveal that, for a commitment with a given takedown proportion, the credit-risk weights present a down sloping put pattern. To wit, for a commitment with a 2-year extension period with 60% funding on the matrix fourth row (namely EP2; p 2 = 0.60), the risk weights decline from $1.565 per $100 of commitment for below investment-grade borrowers (bB ) to $0.898 per $100 of commitment for top investment-grade borrowers (3As to 2As). This pattern is also shown on the three-dimensional Fig. 2, where the X base axis refers to the borrowers risk bucket, the Y base axis to the takedown proportion that varies with the extension duration, and the vertical axis to the banks cost of credit risk per $100 of line commitment. But, when a given risk bucket is chosen, the move in Fig. 2 from the lowest put-value curve to a higher one corresponds to a longer extension period. More concretely, for top credit borrowers in the matrix fifth column, the risk weights increase exponentially from $0.415 for a 1-year straight commitment to $1.504 for a commitment with a 5-year extension period. In brief, both Table 4 and Fig. 2 signal that credit-risk weights, and hence the banks notional costs per $100 of line commitment, are more sensitive to the varying takedown proportion than to the borrowers credit-rating range. Finally, each weight is simply multiplied by 8%, the Cooke ratio, to determine the actual dollar amount of capital per $100 of commitment funding. The computation is summarized in the following stepwise procedure: 1. Compute the sensitivity of the put values, the EPis, to the indebtedness value X and the extension duration T2 T1, respectively,
19 The matrix captures the sensitivity of the extendible put value to the indebtedness value and the extension duration, namely B2EPi / BXB (T 2 T 1). Each row of the resultant matrix is next multiplied by a takedown proportion p i that varies with the duration of the extension period.

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

19

2. Multiply the rows of the resultant matrix by the p i s, the takedown proportions that vary with the duration of the extension period, and 3. Apply the 8% regulatory capital requirement to the new standard credit-risk weights. Or analytically, [EPi = f (X, T 2 T 1 )] [p i = g (T 2 T 1 )] 0.08 = the credit-risk capital charge per $100 of borrower-extendible loan commitments. The same procedure holds also true for straight commitments. 6. Concluding remarks This research makes two contributions. The first one is to price analytically put options and extension premiums embedded in rollover commitments; and the second is to combine extendible put and exercise-cum-takedown proportion when computing the dfairT capital charge corresponding to the commitment dtrueT credit risk. In doing so, the procedure proposes to replace the BIS credit-conversion factor, principal-risk factor, and commitment term-to-maturity dichotomy with three market-based concepts. Namely, (i) the borrower-extendible commitment differentiated on the basis of the duration of the extension period; (ii) an exercise-cumtakedown proportion, and (iii) the value of the put option embedded in the extendible commitment. The fair-value procedure has the advantage that (i) the put value constitutes a finer credit-risk grid than the two artificial values of the conversion and principal-risk factors, and (ii) capital charges computed from risk-weighted balances are quite moderate and internally consistent for all commitment types. Finally, the paper provides new standard commitment risk weights that account for the borrowers rating ranges of public credit agencies. Further work will consider expressing rollover commitments as multiple shout options, as was done for equity options in Windcliff, Forsyth, Vetzal, Verma, and Coleman (2003). For extendible commitments, shout options have the advantage to allow resetting at the end of each commitment or extension period the indebtedness value equal to the line par value; it also allows accounting for multiple extension fees. Another point to elaborate further is the change of markup class by the banks borrower: a matrix of transition probability between mark-up states seems a promising start. Acknowledgment We thank an anonymous referee for insightful comments. We also thank the participants at the Basel meeting of the European Financial Management Association, the Quebec City meeting of the Northern Finance Association, and the 9th Karlsruhe Symposium in Finance, Banking and Insurance for helpful comments and discussions. References
Anathanarayanan, A. L., & Schwartz, E. S. (1980). Retractable and extendible bonds: The Canadian experience. Journal of Finance, 35, 31 47. Andre, P., Matthieu, R., & Zhang, P. (2001). A note on: Capital adequacy and the informational content of term loans and lines of credit. Journal of Banking and Finance, 25, 431 444. Angbazo, L. A., Pei, J., & Sanders, A. (1998). Credit spread in the market for highly leveraged transaction loan. Journal of Banking and Finance, 22, 1249 1282. Athanassakos, G., Carayannopoulos, P., & Tian, P. (1997). Negative option values in extendible Canadian treasury bonds. Advances in Futures and Options Research, 9, 83 100.

20

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

Athavale, A., & Edmister, R. O. (2004). The pricing of sequential bank loans. Financial Review, 39, 231 253. Bank for International Settlements. (1988). International convergence of capital measurement and capital standards. Basel7 Basel Committee on Banking Supervision. Bank for International Settlements. (2000). The standardized approach to credit risk. Basel7 Basel Committee on Banking Supervision. Bank for International Settlements. (2004). International convergence of capital measurement and capital standards: A revised framework. Basel7 Basel Committee on Banking Supervision. Barrios, V. E., & Blanco, J. -M. (2003). The effectiveness of bank capital adequacy regulation: A theoretical and empirical approach. Journal of Banking and Finance, 27, 1935 1958. Berger, A., & Udell, G. (1995). Relationship lending and lines of credit in small firm finance. Journal of Business, 68, 351 382. Black, F. (1976). The pricing of commodity contracts. Journal of Financial Economics, 3, 169 179. Black, F., & Scholes, M. J. (1973). The pricing of options and corporate liabilities. Journal of Political Economy, 81, 637 659. Board of Governors of the Federal Reserve System. (2000). Survey of terms of business lending. Federal reserve board statistical releases E.2. Chateau, J. -P. (1990). Valuation of bcappedQ variable rate loan commitments. Journal of Banking and Finance, 14, 717 728. Chateau, J. -P., & Dufresne, D. (1998). Pricing the put option of bank credit commitments: A two-factor model of credit risk. Schaumburg, IL7 Society of Actuaries. Chateau, J. -P., & Dufresne, D. (2002). The stochastic-volatility American put option of banks credit line commitments: Valuation and policy implications. International Review of Financial Analysis, 11, 159 181. Decamps, J. -P., Rochet, J. -C., & Roger, B. (2004). The three pillars of Basel II: Optimizing the mix. Journal of Financial Intermediation, 13, 132 155. Ergungor, E. O. (2001). Theories of bank loan commitments. Federal Reserve Bank of Cleveland Economic Review, 37, 2 19. Ferguson Jr., R. W. (2003). Capital standards for banks: The evolving Basel accord. Federal Reserve Bulletin, 89, 395 405. Fery, J., Gasbarro, D., Woodliff, D. R., & Zumwalt, J. K. (2003). Market reaction to published and non-published corporate loan announcements. Quarterly Review of Economics and Finance, 43, 1 10. Fisher, Th. (2001). On the new capital adequacy framework based on the second consultative document of the Basel Committee on Banking Supervision. Economic and Financial Review, 8, 181 197. Garman, M. D., & Kohlhagen, S. W. (1983). Foreign currency option values. Journal of International Money and Finance, 2, 231 237. Gottesman, A. A., & Roberts, G. S. (2004). Maturity and corporate loan pricing. Financial Review, 39, 55 77. Hammes, W., & Shapiro, M. (2001). The implications of the new capital adequacy rules for portfolio management of credit assets. Journal of Banking and Finance, 25, 97 114. Hauser, S., & Lauterbach, B. (1996). Empirical tests of the Longstaff extendible warrant model. Empirical Finance, 3, 1 14. Hawkins, G. D. (1982). An analysis of revolving credit agreements. Journal of Financial Economics, 10, 59 81. Ho, T. S. Y., & Saunders, A. (1983). Fixed rate loan commitments, takedown risk and the dynamics of hedging with futures. Journal of Financial and Quantitative Analysis, 18, 499 516. Kashyap, A. K., Rajan, R., & Stein, J. C. (2002). Banks as liquidity providers: An explanation for the coexistence of lending and deposit-taking. Journal of Finance, 57, 33 73. Krahnen, J. P., & Weber, M. (2001). Generally accepted rating principles: A primer. Journal of Banking and Finance, 25, 3 24. Longstaff, F. A. (1990). Pricing options with extendible maturities: Analysis and applications. Journal of Finance, 45, 935 957. Lu, H., 2004. Market value of loan commitments with material adverse change (MAC) covenant. Unpublished research paper, HEC Lausanne. Merton, R. C. (1973). Theory of rational option pricing. Bell Journal of Economics and Management Science, 4, 141 183. Merton, R. C. (1977). An analytic derivation of the cost of the deposit insurance and loan guarantees. Journal of Banking and Finance, 1, 3 11. Morgan, D. P. (1993). Financial contracts when costs and returns are private. Journal of Monetary Economics, 31, 129 146.

J.-P. Chateau, J. Wu / International Review of Financial Analysis 16 (2007) 121

21

Petersen, M. A., & Rajan, R. G. (1994). The benefits of firmcreditor relationships: Evidence from small business data. Journal of Finance, 49, 3 37. Rose, P. S. (2003). Money and capital markets: Financial institutions and instruments in a global market place. New York7 McGraw Hill. Royal Bank of Canada. (2003). Annual report. Toronto7 RBC. Santos, J. A. C. (2001). Bank capital regulation in contemporary banking theory: A review of literature. Financial Markets, Institutions and Instruments, 10, 41 84. Shockley, R. L., & Thakor, A. V. (1997). Bank loan commitment contracts: Data, theory, and tests. Journal of Money, Credit and Banking, 29, 517 534. Thakor, A. V. (1982). Toward a theory of bank loan commitments. Journal of Banking and Finance, 6, 55 84. Thakor, A. V. (1989). Competitive equilibrium with type convergence in an asymmetrically informed market. Review of Financial Studies, 2, 49 71. Thakor, A. V., Hong, H., & Greenbaum, S. I. (1981). Bank loan commitments and interest rate volatility. Journal of Banking and Finance, 5, 497 510. Thakor, A. V., & Udell, G. F. (1987). An economic rationale for the pricing structure of bank loan commitments. Journal of Banking and Finance, 11, 271 289. Windcliff, H., Forsyth, P.A, Vetzal, K. R., Verma, A., & Coleman, T. F. (2003). An object-oriented framework for valuing shout options on high-performance computer architectures. Journal of Economic Dynamics and Control, 27, 1133 1161. Wu, J., 1997. Exotic options and their applications in corporate finance. Unpublished Ph. D. Thesis, Paris.