Empirical estimation of default and asset correlation of large corporates and banks in India

Transcription

1 MPRA Munch Personal RePEc Archve Emprcal estmaton of default and asset correlaton of large corporates and banks n Inda Arndam Bandyopadhyay and Sonal Ganguly Natonal Insttute of Bank Management (NIBM) 16. August 011 Onlne at MPRA Paper No , posted 30. August 011 1:47 UTC

2 Emprcal Estmaton of Default and Asset Correlaton of Large Corporates and Banks n Inda Arndam Bandyopadhyay* Sonal Ganguly August 011 Abstract Estmaton of default and asset correlaton s crucal for banks to manage and measure portfolo credt rsk. Ths would requre studyng the rsk profle of the banks entre credt portfolo and developng the approprate methodology for the estmaton of default dependence. Measurement and management of correlaton rsk n the credt portfolo of banks has also become an mportant area of concern for bank regulators worldwde. The BCBS (006) has specfcally ncluded an asset correlaton factor n the computaton of credt rsk captal requrement by banks adoptng the Internal Ratngs Based Approach. We estmate default correlaton n the credt portfolo of banks. These correlaton estmates wll help the regulator n Inda to understand the lnkage between bank s portfolo default rsks wth the systematc factors. We also derve default and asset correlatons of Indan corporate and compare t wth global scenaro. The work tres to fnd the relatonshp of the correlaton to the default probablty as specfed by the Basel commttee. The fndngs of ths paper could be used further n estmatng portfolo credt rsk, economc captal and rsk adjusted returns on economc captal for large corporate exposures of banks. Key Words: Default Correlaton, Asset Correlaton, Credt Portfolo Rsk JEL Classfcaton: G1, G3, C15 * Address fr correspondence: Dr. Arndam Bandyopadhyay, Assocate Professor, Natonal Insttute of Bank Management (NIBM), Kondhwe Khurd, Pune Emal: arndam@nbmnda.org. Tel

3 1. Introducton. Portfolo rsk s that loss whch arses due to holdng two or more assets n the portfolo. When two or more borrowers default smultaneously, the losses are more severe. The hgher the correlaton of default, the greater s the concentraton portfolo rsk. The lower the correlaton of default more dversfed the portfolo. The crtcal element n successfully managng a credt rsk portfolo s that we must manage the dynamcs of default or asset correlaton rsk. A portfolo approach to credt rsk analyss allows portfolo managers to quantfy and stress test concentraton rsk along varous dmensons. Correlaton descrbes the extent to whch loans tend to default at the same tme. Intutvely, we would expect that companes would have some tendency to default together. Ths jont dependence could happen f the whole economy s n recesson or n dstress; forcng many companes nto bankruptcy at the same tme, or t could be that the default of one company trggers the default of another company. For example, the collapse of a car factory would tend to push supplers and busnesses n the local town closer to default. Correlaton also exsts between companes n dfferent ndustres that rely on the same producton nputs (e.g Car Manufacturng co. s usng Tre) and among companes that rely on the same geographcal market (Damond traders and Textles frms located n Gujarat regon n Inda or IT frms servng the North Amercan market). When correlatons are sgnfcant, they produce loss dstrbutons that are hghly skewed (tal measures of credt rsk lke value at rsk captures ths). Modelng credt qualty correlaton and default correlaton s therefore crucal for banks to manage and measure portfolo credt rsk. Ths would requre studyng the rsk profle of the banks entre credt portfolo and developng the approprate methodology for the estmaton of default dependence. Measurement and management of correlaton rsk n the credt portfolo of banks has also become an mportant area of concern for bank regulators worldwde. The BCBS (006) has specfcally ncluded an asset correlaton factor n the computaton of credt rsk captal requrement by banks adoptng the Internal Ratngs Based Approach. Internatonally, several methodologes have been developed to estmate default correlaton and several authors have documented the relatonshp between the ntal credt qualty of the portfolo and the default and asset correlaton for commercal portfolos. The 1

4 structural model approach uses equty correlaton as a proxy for asset correlaton. Ths approach s based on the work by Merton (1974), accordng to whch loan default occurs when the market value of the frms assets falls below the book value of debt. Thus the default correlaton between two borrowers s constructed wth the use of the correlaton of the borrower s asset returns (derved from equty returns) and the normal nverse to the dstance to default. De Servgny and Renault (00) use a sample of 1101 frms from S&P s 1 ndustry categores and calculate average equty correlatons across and wthn ndustres over the perod Bluhm and Overbeck (003) have deduced a methodology for estmaton of asset correlatons based on the volatlty of default rates tme seres. Usng Moody s bond default data from 1970 to 001 they derve grade-wse mplct asset correlaton of US corporates. They emprcally observed that ther estmated asset correlaton s n lne wth Basel II prescrbed asset correlaton for corporates. Lopez (004) has used the structural model framework to emprcally derve asset correlaton for portfolos. Hs paper demonstrates that asset correlaton for relatvely hghly rated, large szed companes s hgh. Accordng to hs explanaton, ths relatonshp arses because hgh credt qualty frms are more lkely to be nfluenced by common macroeconomc condtons. On the other hand, asset correlaton for poor credt qualty, large szed companes are low because defaults of such frms are subject to frm-specfc problems. Ths s also the relatonshp whch s hghlghted by BCBS (006) wheren, n the credt rsk captal estmaton formula for large corporate exposures, asset correlaton s a decreasng functon of probablty of default. Bandyopadhyay, Saha and Chherawala (007) use the approach developed by Lucas (1995) to establsh a lnkage between ratng wse PD and Default Correlaton amongst Indan corporates. We estmate default correlaton n the credt portfolo of banks as well as for rated corporates n Inda. These correlaton estmates wll help the regulator n Inda to understand the lnkage between bank s portfolo default rsks wth the systematc factors. We also derve asset correlatons of Indan corporate and compare t wth global scenaro. The work tres to fnd the relatonshp of the correlaton to the default probablty as specfed by the Basel commttee. Ths paper demonstrates that default correlaton s postvely related to the default probablty of frms. The fndngs of ths paper could be used further n estmatng portfolo

5 credt rsk, economc captal and rsk adjusted returns on economc captal for large corporate exposures of banks. Ths paper s organzed as follows. Next secton descrbes the data used n the study and methodology followed by us. Secton 3 dscusses the results and ther nterpretaton. The fnal secton 4 addresses the major conclusons of the paper.. Data and Methodology The frst part of ths project that deals wth default correlaton n the credt portfolo of 15 major banks on Inda s based on the Non Performng Assets (NPA) movements data on ndvdual banks. The bank level advances and NPA data was taken from Centre for Montorng Indan Economy (CMIE) Prowess. Here we use sngle default correlaton on the banks. The second part of the paper deals wth estmatng the default and asset correlaton on Indan corporate ratng wse. We have used CRISIL s publshed yearly bond ratng data of 57 large corporates and studed ther hstorcal ratng mgraton pattern from 1993 to 009. Both default and asset correlaton has been estmated for Moody s one-year default rates ( ). The methodology has been descrbed below: Consder a smple case of a portfolo of loans. The unexpected loss (UL) for ths portfolo s gven by the varance equaton: UL P = ρ UL + ρ UL + ρ UL UL Unexpected loss (loss volatlty) of a loan can be expressed as: 1 UL = EAD LGD PD ( 1 PD) Where, EAD s the average exposure at default, LGD s the average loss gven default and PD s the probablty of default. If we consder a large number of loans n the pool, we can the portfolo UL p wll be: UL P = N N = 1 j= 1 ρ UL UL (1) j j We can get an estmate for the correlaton of default f we assume that the correlaton between each loan s dentcal (.e. ρ = ρ j ): 3

6 UL P = N N = 1 j= 1 ρ UL UL j j N N = ρ = 1 j= 1 UL UL j Assumng each loan has the same UL (.e. UL =UL j ), we can estmate the correlaton as follows: UL ρ = P () ( N UL ) Thus, usng ths concept, we have calculated the default correlaton n the credt portfolo of 15 Indan banks. Ths correlaton s a measure for the senstvty of the Bank s ncremental rsk of default of loans to the systematc factors whch represents the state of the economy. We apply the same methodology to calculate the default correlaton of Moody s oneyear default rates and to the ratng wse data of Indan corporate. These correlatons measure the senstvty of the ratng grade s PD to the macro economc (or systematc) factors. An alternatve approach to dervng emprcal default correlatons s proposed by Gordy and Hetfeld (00) usng a factor model of credt rsk. Credt Metrcs (1997), Crouhy et al. (000) and Zhang et al. of MKMV (008) derve asset return correlatons from a structural model whch lnks correlatons to fundamental factors. Our next step s to calculate the asset correlaton wthn each ratng n the global as well as natonal data. For ths we used the varance equaton gven by Bluhm, Overbeck & Wagner (003, Basel II Handbook) derved from a factor model: Varance of condtonal default rates g(y) s expressed by: V [ g( y)] = JDP( PD, ρ ) PD PD (3) Where, PD = mean value of tme seres Varance = sample varance of tme seres of observatons of default rates = default rate volatlty mpled asset correlaton JDP s the jont default probablty of two oblgors n a unform portfolo wth parameters probablty of default PD and mplct asset correlaton. The one factor rsk model uses the computaton of the asset correlaton followng the Merton Model. In ths, a default event occurs f the frm value of oblgor crosses the default 4

7 threshold. The default of an oblgor s drven by a latent varable whch s a functon of a systematc factor and a frm specfc dosyncratc factor. We start by assumng that the normalzed asset return R t of a frm n the credt portfolo s drven by common macro factor y and an dosyncratc factor. Therefore, the asset returns at a chosen horzon (say 1 year) can be wrtten as: R = ρ y + 1 ρ ε (4).. d.. d y ~ N(0,1); ε ~ N(0,1) R s the mplct return on frm s asset that s drvng default mgraton. var( y) = var( ε ) = var( R ) = 1 Strong senstvty to the systematc factor mplcates a hgher correlaton of the borrower and hgher volatlty of the default rates. Note that default correlaton and asset correlaton are not the same. Default correlaton estmate s much lower than the asset correlaton. The mplct asset values of two oblgors at the horzon are jontly normally dstrbuted and ther JDP follows a bvarate dstrbuton. The jont default probablty can therefore be obtaned by usng the followng expresson: JDP = Pr[ A K, A K ) = N( K j j j, K j, ρ) Where, N(.) denotes the cumulatve bvarate standard normal dstrbuton of the followng form: 1 JDP = π 1 ρ N ( PD ) N ( PD ) exp 1 x ρx x j 1 ρ + x j dxdx The two upper ntegraton lmts refer to the default thresholds (K,K j ) for a homogenous portfolo. It s assumed that asset returns are normally dstrbuted a pror. Symbol ρ represents the asset correlaton between two oblgors; x and y are ther default thresholds. We can calculate the asset correlaton a f we already know the JDP by usng the followng formula: JDPj = N( K, K j, ρ a ) j (5) 5

8 Where N denotes the cumulatve bvarate standard normal dstrbuton K, Kj gves the dstance to default a s the asset correlaton We calculate JDP usng the BIVNOR functon such that JDP j = BIVNOR (normsnv(pd ), normsnv(pd j ), a ) (5a) BIVNOR s a functon that gves the cumulatve bvarate standard normal dstrbuton. Gordy used smlar measure to calculate JDP. The next step s to estmate mplct asset correlaton through teratons. The JDP value s the key ngredent of the formula expressng the varance of the condtonal default rates g(y) as depcted n expresson 3. We estmate the average default rates PD from the hstorcal tme seres data. V[g(y)] s estmated by the sample varance of the tme seres of observed default rates. Fnally, we substtute these to nputs n equaton 3 to obtan the only unknown parameter = a through optmzaton method. Thus, solvng for yelds default rate volatlty mpled asset correlaton a. Ths correlaton generally captures the changng macro economc scenaro (or the systematc factors). Correlaton or dependence s captured by the varatons n default rates due to macro economc changes. Followng ths approach, we have estmated ratng class wse mplct asset correlaton based on Moody s and CRISIL data. 3. Emprcal Results and Interpretaton Table 1 shows the PD, LGD and correlaton fgures of the 15 banks. Ths table s based on the bank data (NPA movements) as reported n CMIE Prowess. Here we fnd that Canara bank has the hghest default correlaton whle ICICI bank has the lowest. 6

9 Table-1- Default Correlaton of Indan banks BANKS PD LGD Portfolo UL Correlaton Allahabad Bank 1.97% 48.53% 0.49% 0.53% Andhra Bank 1.1% 44.81% 0.68%.07% Bank of Baroda 1.51% 53.69% 0.74% 1.9% Bank of Inda 1.91% 44.10% 0.69% 1.30% Canara Bank.44% 8.08% 1.06% 5.93% Central Bank of Inda.19% 65.3% 1.06% 1.3% Corporaton Bank 1.06% 58.86% 0.43% 0.5% Dena Bank 3.14% 43.97% 1.1%.14% HDFC Bank.43% 63.99% 1.% 1.54% ICICI Bank 1.65% 64.37% 0.4% 0.6% Indan Overseas Bank.64% 4.60% 1.08%.49% Punjab Natonal Bank 1.9% 45.8% 0.53% 0.71% State Bank of Inda.54% 49.67% 0.74% 0.90% UCO Bank.7% 48.75% 0.79% 1.19% Unon Bank of Inda 1.73% 56.3% 0.47% 0.41% The table shows the one-year default rates ratng wse for a perod The default and asset correlaton has been calculated accordng to the methodology descrbed n the prevous secton. The default correlaton calculated ncreases wth ncreasng default rates and thus t s found to be hghest for Caa-C category and lowest for Aaa category. However an nterestng pont to note here s that A category has lower default correlaton than Aa category. Accordng to the data the overall PD of A category s also lower than that of Aa category. 7

10 Table-- Default and Asset Correlaton usng Moody s one-year default rates Aaa Aa A Baa Ba B Caa-C % 0.000% 0.000% 0.541% 4.64% % % % 0.000% 0.000% 0.000% 0.881% 0.000% 14.86% % 0.000% 0.000% 0.000% 0.000% 7.73% % % 0.000% 0.000% 0.460% 0.000% 3.9% % % 0.000% 0.000% 0.000% 0.51% 7.143% 0.000% % 0.000% 0.000% 0.000% 1.00% 6.154% 0.000% % 0.000% 0.000% 0.000% 0.985% 0.000% 0.000% % 0.000% 0.000% 0.91% 0.55% 3.175% % % 0.000% 0.000% 0.000% 1.090% 5.556% 0.000% % 0.000% 0.000% 0.000% 0.494% 0.000% 0.000% % 0.000% 0.000% 0.000% 0.000% 4.938% % % 0.000% 0.000% 0.000% 0.000% 4.494% 0.000% % 0.000% 0.56% 0.318%.783%.99% 3.077% % 0.000% 0.000% 0.000% 0.911% 6.306% 4.105% % 0.000% 0.000% 0.36% 0.833% 6.780% % % 0.000% 0.000% 0.000% 1.413% 7.483% 0.000% % 0.000% 0.000% 1.347%.041% 11.60% 6.667% % 0.000% 0.000% 0.000%.74% 6.154% 0.000% % 0.000% 0.000% 0.000% 1.58% 6.0% 8.571% % 0.604% 0.000% 0.594% 3.037% 8.70% 5.000% % 0.000% 0.000% 0.000% 3.409% % 58.84% % 0.000% 0.000% 0.7% 4.89% 1.598% % % 0.000% 0.000% 0.000% 0.306% 9.18% 8.571% % 0.000% 0.000% 0.000% 0.567% 4.517% 6.667% % 0.000% 0.000% 0.000% 0.4% 4.050% 5.63% % 0.000% 0.000% 0.000% 0.714% 4.7% 9.091% % 0.000% 0.000% 0.000% 0.000% 1.366% % % 0.000% 0.000% 0.000% 0.188% 1.935% % % 0.000% 0.000% 0.118% 0.808% 3.780% 11.55% % 0.000% 0.000% 0.103% 1.61% 4.967% 18.18% % 0.000% 0.000% 0.378% 0.859% 5.803% 0.073% % 0.000% 0.165% 0.191% 1.308% 9.501% % % 0.000% 0.166% 1.8% 1.481% 4.574% 8.19% % 0.000% 0.000% 0.000% 0.955%.073% 1.016% % 0.000% 0.000% 0.000% 0.381% 0.83% % % 0.000% 0.000% 0.175% 0.000% 1.4% 6.179% % 0.000% 0.000% 0.000% 0.181% 1.144% 5.919% % 0.000% 0.000% 0.000% 0.000% 0.000% 5.873% % 0.515% 0.333% 0.454% 1.058% 1.985% 14.53% Overall PD 0.000% 0.09% 0.04% 0.175% 1.11% 5.341%.055% ULP 0.000% 0.15% 0.074% 0.317% 1.19% 4.377% 0.409% ULT 0.000% 1.694% 1.536% 4.18% %.485% 41.46% DC % 0.34% 0.574% 1.9% 3.789% 4.30% JDP % % % 0.066% % 9.096% AC -- 7% 0% 16% 1% 14% 4% Note: PD: Long term average Probablty of Default; ULP s the portfolo volatlty (or unexpected loss); ULT s the total portfolo unexpected loss assumng perfect correlaton; DC=Default correlaton; JDP: Jont default probablty; AC: Implct Asset correlaton. 8

11 The asset correlaton computed gves us nterestng results. Whle t s stll hghest for Caa-C category and lowest for Aaa category, the asset correlaton for Aa category s hgher than that of A, Baa, Ba and B categores. Table-3- Default Correlaton of Indan corporate (ratng wse) AAA AA A BBB BB B C IG NIG % 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% % 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% % 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% % 0.00% 0.00% 1.85% 0.00% 0.00% 0.47% 16.67% % 0.00%.33% 0.00% 18.18% 0.00% 1.13% 16.67% % 0.00% 1.54% 0.00% 3.53% 50.00% 50.00% 0.7% 30.43% % 1.35% 7.1% 18.18% 55.17% 0.00% 50.00% 6.17% 50.00% % 0.00% 0.00% 5.71% 6.67% 4.86% 57.14% 0.97% 34.09% % 0.00% 1.69% 8.70% 1.50% 16.67% 50.00% 1.66% 5.00% % 0.00% 0.00% 11.76%.% 60.00% 33.33% 1.18% 35.00% % 0.00% 3.70% 0.00% 40.00% 0.00% 66.67% 0.64% 36.36% % 0.00% 0.00% 14.9% 0.00% 0.00% 0.00% 0.68% 0.00% % 0.00% 0.00% 0.00% 0.00% 100.0% 0.00% 66.67% % 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% % 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% % 0.00% 0.00% 0.00% 100.0% 0.00% % % 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% Average 0.00% 0.08% 0.97% 3.56% 18.7% 4.% 31.3% 0.80% 4.17% PD ULP % 1.94% 5.97% 6.61% 6.17% 33.63% 1.48% 8.3% ULT --.8% 9.80% 18.53% 39.01% 4.84% 46.38% 8.91% 4.81% DC % 3.93% 10.37% 46.54% 37.3% 5.57%.77% 43.75% In the Indan context too, default correlaton ncreases wth ncreasng default rates. However the default correlaton of category B s lower than that of BB category. 9

12 Table-4- Asset Correlaton of Indan corporate (ratng wse) Ratng Avg PD Var(PD) JDP Implct AC IG 1.0% 0.0% 0.03% % NIG % 8.019% % % AAA 0.000% 0.000% 0.000% 0.000% AA 0.097% 0.001% 0.00% % A 1.618% 0.038% 0.064% % BBB 3.56% 0.356% 0.483% % BB 6.761% 7.08% 14.43% % B 9.630% 6.848% 15.68% 51.50% C % % 31.7% % Here, we have calculated the asset correlaton of the Indan corporate (ratng wse) usng the varance equaton. Here too n general we can say that the asset correlaton ncreases wth ncreasng default rates. However t s not true for A category whch shows the lower asset correlaton than category AA. Frst, let us analyze the default correlaton of the banks. The chart below compares the default correlaton among the 15 banks Chart 1 10

13 Correlaton captures the systematc rsk. More granular the portfolo, hgher the dversfcaton and thus lower the default correlaton. From our data we can say that Canara Bank has the hghest default correlaton whle ICICI has the lowest. Ths gves us an estmate about the probablty of all the loan assets n a partcular banks credt portfolo to default together. Ths fgure s more mportant durng tmes of downturn. If we compare the fgures of ULp and ULt for each bank, we are able to analyze the correlaton better. The default correlaton compares the actual unexpected loss of the portfolo (ULp) to the sum of the ndvdual unexpected loss of all the assets n the portfolo (ULt). Due to dversfcaton benefts, ULp s always lower than ULt. Hgher the ULp, that means lower the dversfcaton n the portfolo and thus hgher the correlaton. Chart compares the ULp and ULt fgures of the banks. If we take the example of Canara Bank, the ULt s qute low (4.33%) but the ULp(1.06%) s among the hghest n the group of 15 banks. If we compare ths wth Allahabad Bank, the ULt s very hgh (6.75%), but the ULp,.e, the actual UL of the portfolo s among the lowest at 0.49%. ths shows that Allahabad Bank has a hghly dversfed portfolo. 11

14 Chart- Next, we estmated the default correlaton on Moody s one year default rates ratng wse. The chart below compares the default correlaton wth the overall PD for each ratng. A general observaton here s that the default correlaton follows the same trend as that of the overall PD. We fnd from the data that the overall PD for A category s lower than that of Aa. Smlarly, the default correlaton too follows the same trend,.e, t ncreases wth decreasng credt ratng but s lower n case of A than that of Aa. That means PD s the man drvng force for default correlaton,.e, wth ncreasng PD, default correlaton also ncreases. Chart- 3 further elaborates ths. 1

15 Chart-3 The same trend s seen n Indan corporate (Chart-4). As the ratng grade worsens,.e, the PD ncreases, the default correlaton also ncreases. Default correlaton for B category s shown lower than BB, but ths may be due to lack of adequate data. Chart 4 Ths result s of sgnfcance for Indan banks snce t mples that poor credt qualty commercal loan portfolos would have to be supported by a hgher level of economc captal not just because the default probabltes n such portfolos wll be hgh but also because of hgher nherent default correlatons between poor credt qualty borrowers. Next we estmate the asset correlaton usng varance equaton. The asset correlaton shows how the asset value of one borrower depends on the asset value of another. Lkewse t can be descrbed as the dependence of the asset value of a borrower on the general state of the economy. Asset correlatons are also an mportant component of the Basel II Accord for regulatory captal requrements of credt rsk portfolos. In the Basel Commttee on Bankng Supervson (BCBS) document of June 006 asset correlatons for soveregns, banks and corporates were prncpally assumed to be between 1% and 4%, dependng on the probablty of default (PD) assumng that asset correlaton declnes wth PD percentage. For the lowest PD borrower the asset correlaton s 4% and for the hghest PD the asset correlaton s 1%.We note that for small and medum szed corporates an extra downward frm-sze adjustment up to 4% s made and ths brngs the effectve range of corporate asset correlatons between 8% and 4%. 13

16 Chart-5 shows the asset correlaton based on Moody s one-year default rates for a perod from An nterestng pont to note here s the negatve relatonshp between PD and asset correlaton. Ths fndng corroborates what had been stated by the BCBS (006). Chart-5 Chart 6 shows that the same trend s followed n case of Investment grade and speculatve grade. An mportant argument for ths negatve relaton between asset correlaton and PD s that, as the frm approaches default, ts frm specfc rsk ncreases and ts systematc rsk (gven by asset correlaton) decreases. However, we dd not see the same trend n the Indan corporate data. Ths may be due to lack of adequate data n Inda. 14

17 Chart-7 Whle n case of nvestment and non-nvestment grade, the asset correlaton ncreases wth ncreasng PD, there s no partcular trend vsble n the corporate ratngs. 4. Conclusons The frst part of ths paper derves the estmates of default correlaton n the credt portfolo of Indan banks (publc and prvate sector) and fnds ts relaton to the default probablty. The results substantate the presence of default correlaton due to macroeconomc factors and/or macro-economc factors. Snce default events are not ndependent, as s evdent from our fndngs, the correlaton effects need to be consdered carefully n managng and measurng the concentraton rsk n credt portfolos. The second part of the paper deals wth the asset correlaton fgures. There s found to be a negatve relatonshp between asset correlaton and the probablty of default. Ths relatonshp, as prescrbed by Basel II IRB document, has been found n the fndngs based on Moody s data. However, we do not fnd any smooth relatonshp between the probablty of default (PD) and asset correlaton for Indan corporate. The asset correlaton range s also dfferent n comparson to what s prescrbed for corporate exposures by BCBS (1% to 4%). These fndngs have large mplcatons for banks. Asset correlaton factor has been ncluded n the Basel II IRB approach for calculaton of rsk weghts. Rght now, banks gve very lttle attenton to the Pllar II rsk, as could be seen from the hgh default correlaton 15

18 fgures for some banks. However f not taken care of, the correlaton rsk can prove to be a major reason for losses n downturn. Reference: Bandyopadhyay, A. and Chherawala, T. and Saha A. (007), Calbratng asset correlaton for Indan corporate exposures: Implcatons for regulatory captal, The Journal of Rsk Fnance, Vol. 8, No. 4, pp Basel Commttee on Bankng Supervson (BCBS), (006), Internatonal Convergence of captal Measurement and Captal Standards: A Revsed Framework, Publcaton No. 18, Bank for Internatonal Settlements, Basel, June. Bluhm, C. and L. Overbeck, (003), Systematc rsk n homogeneous credt portfolos ; n: Credt Rsk; Measurement, Evaluaton and Management n; G. Bol et al (eds), Contrbutons to Economcs (Hedelberg, Physca: Verlag/Sprnger). Bluhm, C. and L. Overbeck, (007), Explanng the correlaton n Basel II: Dervaton and evaluaton, n Mchael Ong (eds), The Basel Handbook: A gude for fnancal practtoners (RISK books). Credt Metrcs. (1997), Techncal document, JP Morgan. Crouhy, M. and Gala, D. and Mark, R. (000), A comparatve analyss of current credt rsk models, Journal of Bankng & Fnance, Vol. 4, Nos. 1-,, pp Frey, R. and McNel, A. J. and Nyfeler, M. A. (001), Modellng dependent defaults: Asset correlatons are not enough! Workng paper, ETH Zurch. Gordy, M. B. (000), A comparatve anatomy of credt rsk models, Journal of Bankng & Fnance, Vol. 4, ssues 1-, pp Lopez, J. A. (004), The emprcal relatonshp between average asset correlaton, frm probablty of default, and asset sze, Journal of Fnancal Intermedaton, Vol. 13, No., pp Lucas, D. J. (1995), Default correlaton and credt analyss, Journal of Fxed Income, Vol. 4, No. 4, pp Merton, R. (1974), On the prcng of corporate debt: The rsk structure of nterest rates, Journal of Fnance, Vol. 9, pp

19 Moody s Investors Servces (009), Corporate Default and Recovery Rates, , February. Zhang J. and Zhu, F. and Lee, J. (008), Asset correlaton, realzed default correlaton, and portfolo credt rsk, MKMV workng paper. March 3. 17