New Economic School, 2005/6

FINANCIAL RISK MANAGEMENT

Lecture notes Plan of the course Classification of risks and strategic risk management Deri ati es and financial engineering !arket risk Li"uidit# risk Credit risk $%erational risk

Lecture 1. Introduction
Plan Definition of risk !ain t#%es of risks E&am%les of financial failures S%ecifics of financial risk management Em%irical e idence on '! %ractices hat is ris!" Chinese h#eroglif (risk) o Danger or o%%ortunit# o *his is the essence of financial risk+management, -ncertaint# s risk o Su./ecti e / o./ecti e %ro.a.ilities o S%eculati e / %ure #o$ to %easure ris!" 0ro.a.ilit# / magnitude / e&%osure o S#stematic s residual risk !a&imal s a erage losses 1.solute s relati e risk #o$ to classif& ris!s" Nature / %olitical / trans%ortation / 2 Commercial o 0ro%ert# / %roduction / trade / 2 o 3inancial 4n estment5 lost o%%ortunit# 6e7g7 due to no hedging8, direct losses, lower return 0urchasing %ower of mone#5 inflation, currenc#, li"uidit# Main t&'es of financial ris!s !arket risk o 4nterest rate / currenc# / e"uit# / commodit#

3inancial 'isk !anagement, NES 2006/9

Credit risk o So ereign / cor%orate / %ersonal Li"uidit# risk o !arket / funding $%erational risk o S#stem : control / management failure / human error E ent risk

E(a%'les of financial failures

Lessons for ris! %ana)e%ent 4ntegrated a%%roach to different t#%es of risks 0ortfolio iew 1ccounting for deri ati es !arket microstructure 'ole of regulators and self+regulating organi;ations Methods for dealin) $ith uncertaint& *& Kni)ht Consolidation S%eciali;ation Control of the future 4ncreased %ower of %rediction Financial ris! %ana)e%ent 1 oid< o =ut #ou cannot earn mone# without taking on risks 'eser es5 es%7 .anks Di ersification o =ut5 onl# nons#stematic risk >edging o -suall#, using deri ati es 4nsurance5 for e&ogenous low+%ro.a.ilit# e ents 2

3inancial 'isk !anagement, NES 2006/9

o $therwise .ad incenti es E aluation .ased on risk+ad/usted %erformance Strategic '!5 enter%rise+wide %olic# towards risks o 4dentification / !easurement / !anagement / !onitoring

Should the co%'anies hed)e" N+ *he !! irrele ance argument o *he firm?s alue is determined .# its asset side *he C10! argument o @h# hedge uns#stematic 6e7g7, 3A8 risk< o 1n# decrease in B will .e accom%anied .# decrease in ECC3D, *ransactions with deri ati es ha e negati e e&%ected alue for a com%an# o 1fter fi&ed costs Should the co%'anies hed)e" ,ES =oth !! and C10! re"uire %erfect markets o =ankru%tc# costs are im%ortant *he C10! re"uires di ersification o 'eal assets are not er# li"uid and di isi.le Shareholder wealth ma&imi;ation o !arket frictions5 financial distress costs / ta&es / e&ternal financing costs !anagerial incenti es o 4m%ro ing e&ecuti e com%ensation and %erformance e aluation 4m%ro ing decision making E%'irical e-idence on RM 'ractices 3inancial firms5 o *he si;e of deri ati e %ositions is much greater than assets 6often more than E0 times8 Non+financial firms5 o !ain goal5 sta.ili;e C3s o 3irms with high %ro.a.ilit# of distress do not engage in more '! o 3irms with enhanced in o%%ortunities and lower li"uidit# are more likel# to use deri ati es Methods for dealin) $ith uncertaint& *& Kni)ht Consolidation S%eciali;ation Control of the future 4ncreased %ower of %rediction Financial ris! %ana)e%ent 1 oid< o =ut #ou cannot earn mone# without taking on risks 'eser es5 es%7 .anks Di ersification o =ut5 onl# nons#stematic risk >edging o -suall#, using deri ati es 4nsurance5 for e&ogenous low+%ro.a.ilit# e ents F

3inancial 'isk !anagement, NES 2006/9

o $therwise .ad incenti es E aluation .ased on risk+ad/usted %erformance Strategic risk management5 enter%rise+wide %olic# towards risks o 4dentification / !easurement / !anagement / !onitoring

Current trends in ris! %ana)e%ent Deregulation of financial markets 4ncreasing .anking su%er ision and regulation *echnological ad ances 'esults5 risk aggregation, increasing s#stemic and o%erating risks

3inancial 'isk !anagement, NES 2006/9

Lecture .. Financial en)ineerin)

Plan S%ecifics of risks of different instruments o 4n estment strategies / %ricing / s#stematic risks Stocks / .onds / deri ati es o 3orwards / futures / o%tions / swa%s General a''roach to financial ris! %odelin) -se of returns o Stationar# 6in contrast to %rices8 'isk ma%%ing5 %ro/ecting our %ositions to 6a small set of8 risk factors o @e might not ha e enough o.ser ations for some %ositions E7g7, new market or instrument o *oo large dimensionalit# of the co ariance matri& 3or n assets5 n ariances and n6n+E8/2 correlations o E&cessi e com%utations during simulations S'ecifics of ris!s for different assets Discounted cash flow a%%roach5 00 H It C3t/6EJr8t Stocks5 00 H 60EJDi E8/6EJr8 H ItHE5K Di t/6EJr8t o 4nterest rates / E&change rates o 0rices on goods and resources o Cor%orate go ernance / 0olitical risk =onds5 00 H ItHE5* C/6EJrt8t J 3/6EJr*8* o 4nterest rates for different maturities o Default risk Deri ati es o 0rice of the underl#ing asset Sha%e of the %a#off function Lolatilit# o 4nterest rates Inde( %odels5 'i,t H Mi J IkNki4kt J Oi,t, k where E6Oi,t8H0, co 64 , Oi8H0, and E6OiO/8H0 for iP/ 'isk management5 Q'i R IkNkiQ4k Se%aration of total risk on s#stematic and idios#ncratic5 ar6'i8 H Ni2S2! J S26O8i o S#stematic risk de%ends on factor e&%osures 6.etas85 Ni2S2! o 4dios#ncratic risk can .e reduced .# di ersification Co ariance matri&5 co 6'i, '/8 H NiN/S2! o Correlations com%uted directl# from the historical data are .ad %redictors

Stocks
Single+inde& model with market factor5 'i,t H Mi J Ni'!t J Oi,t 6Market model, if we don?t make an assum%tion E6OiO/8H0 for iP/8 where NHco 6'i, '!8/ ar6'!85 6market8 .eta, sensiti it# to the market risk !ulti+inde& models5 5

3inancial 'isk !anagement, NES 2006/9

4ndustr# indices !acroeconomic factors o $il %rice, inflation, e&change rates, interest rates, TD0/ consum%tion growth rates 4n estment st#les o Small+ca% / large+ca% Salomon Brothers: o Lalue / growth 6low/high 0/E8 -Economic growth o !omentum / re ersal -Default spread Statistical factors -LR interest rates o 0rinci%al com%onents -SR interest rates

In-est%ent strate)ies -US dollar S%eculati e5 choosing higher .eta o 4ncreases e&%ected return and risk o -sed .# more aggressi e mutual funds >edging 6s#stematic risk85 N H 0 o !arket+neutral strateg#5 return does not de%end on the market mo ement o $ften used .# hedge funds 1r.itrage5 riskless %rofit 6(free lunch)8 o =u# under alued asset and sell o er alued asset with the same risk characteristics o 0ure ar.itrage is er# rare5 there alwa#s some risks

-Inflation shock

=onds
Single+inde& model with interest rate5 'i,t H ai J Di Q#t J ei,t where #t5 interest rate in %eriod t, D5 duration, e&%osure to interest risk !acaule# duration5 D H +CU0/0D/CU#/#D H +ItHE5* t Ct / 60 #t8 3or the .ond with the %rice5 00 H ItHE5* Ct / #t @td+a erage maturit# of .ond %a#ments, D V * Elasticit# of the .ond?s %rice to its W*! 6#ield to maturit#8 3or small changes in B5 Q0/0 R +D Q#/# H +DX Q# o DX H D/#5 modified duration Con e&it#5 C H +ItHE5* t6tJE8Ct / 60 #tt8 3or small changes in B5 Q0/0 R +D Q#/# J Y C 6Q#/#82 Asset/lia*ilit& %ana)e%ent5 used .# %ension funds, insurance com%anies Ta% anal#sis5 ga%t H 1t+Lt o 0ositi e ga% im%lies higher interest income in case of rising B 0erfect hedging5 ;ero ga%s 6cash flow matching8 o Can .e unachie a.le or too e&%ensi e 4mmuni;ation5 D6assets8 H D6lia.ilities8 o 1cti e strateg#, since .oth duration and the term structure of interest rates e ol e o er time o Need %recise measure of duration 6and con e&it#8 o Does not %rotect from large changes in B

Deri ati es
0eri-ati-es5 6

3inancial 'isk !anagement, NES 2006/9

-n.undled contingent claims o 3orwards / 3utures / Swa%s / $%tions Em.edded o%tions5 o Con erti.le / redeema.le .onds 'ole of deri ati es5 efficient risk sharing o S%eculation5 gi e high le erage o >edging5 reduce undesira.le risks Notional si;e5 around ZEG0 trln o *wice as large as e"uit# and .ond markets com.ined *he total market alue 6.ased on %ositi e side85 less than ZF trln

For$ard 1 futures $.ligation to .u# or sell the underl#ing asset in %eriod * at fi&ed settlement %rice [ \ero alue at the moment of signing the contract 6tH08 0a#off at *, long %osition5 S*+3 For$ard S%ecific terms S%ot settlement Low li"uidit# o !ust .e offset .# the counter deal Credit risk Futures Standardi;ed e&change+traded contract o 1mount, "ualit#, deli er# date, %lace, and conditions of the settlement Credit risk taken .# the e&change o *he e&change clearing+house is a counter+%art# o Collateral5 the initial / maintenance margin o !arking to market dail# Long %osition5 recei e 163t+3t+E8 into account >igh li"uidit#, %o%ular among s%eculators o Can .e offset .# taking an o%%osite %osition o -suall#, cash settlement No/ar*itra)e for$ard 'rice 3 6assuming %erfect markets85 3or assets with known di idend #ield 5 3 H Se6r+"8* o Lalue of the long %osition5 63+[8e+r* H Se+"* +[+r* S&ste%atic ris!s Delta 6first deri ati e wrt the %rice of the underl#ing85 ]He+"* Tamma 6second deri ati e wrt the %rice of the underl#ing85 ;ero, S'ecifics of futures 4f rHconst, futures %rice H forward %rice 4f r is stochastic and corr6r, S8^0, futures %rice ^ forward %rice o *he margin %roceeds will .e re+in ested at higher rate Li"uidit# risk due to margin re"uirements 9

3inancial 'isk !anagement, NES 2006/9

=asis risk5 the !asis H s%ot %rice _ futures %rice o 4deal hedge5 the .asisH0 at the deli er# date o -suall#, the .asis ^ 0 at the settlement date !aturit# / "ualit# / location risks

E(a%'le2 !etallgesellschaft Sold a huge olume of 5+E0 #ear oil forwards in E``0+`F, hedging with short+term futures @hen the oil %rice fell, the margin re"uirements e&ceeded ZE .ln7 *he =oard of Directors decided to fi& the futures? losses and close forward %ositions7 *he final losses were around ZE7F .ln7 Lessons5 o *he rollo er .asis risk was ignored .# those managers who designed the strateg# o *he senior management did not understand this strateg# and therefore made clearl# inefficient decision to close long forward %ositions that were %rofita.le after decline in oil %rices7 In-est%ent strate)ies S%eculati e o Naked5 .u#ing or selling futures o S%read5 calendar / cross >edging o E7g7, short hedge5 we need to sell the underl#ing asset, hedge with short futures >edge ratio5 hedged %osition / total %osition o >edging stock e&%osure with stock inde& futures5 NS o >edging interest rate risk with duration5 immuni;ation +'tions5 Euro%ean call 6%ut85 right to .u# 6sell8 the underl#ing asset at the e&ercise date * at the strike/e&ercise %rice [ 1merican call 6%ut85 can .e e&ercised at an# time .efore * 'ight, no o.ligation 6for the .u#er8 H^ as#mmetric %a#off function o Call5 c* H ma&6S*+[, 08 o 0ut5 %* H ma&6[+S*, 08 S#nthetic forward5 long call, short %ut Euro%ean call+%ut %arit#5 c0 J [e+r* H %0 J S0 o Co ered %ut H call J cash S'eculati-e strate)ies Naked / co ered o%tion S%read5 o%tions of one t#%e o =ear / .ull5 long and short call 6%ut8 o =utterfl#5 long with [E and [F, two short with [2H Y 6[EJ[F8 o Calendar5 short with * and long with *Jt with the same strike Com.ination5 o%tions of different t#%e o Straddle5 call and %ut o Stri%5 call and two %uts o Stra%5 two calls and %ut o Strangle5 with different strikes 3lac!/Scholes %odel Call5 ct H Ste+"* N6dE8 _ Ae+r* N6d28 a

3inancial 'isk !anagement, NES 2006/9

0ut5 % H Ae+r* N6+d28 _ Se+"* N6+dE8 o dE H Cln6S/A8 J *6r+"JS2/28D / CSb*D, d2 H dE _ Sb* o is cont7 di idend #ield o N678 is a std normal distri.ution function Ti en %rice, S is im%lied olatilit# o Tood forecast of future olatilit# of the underl#ing

S#stematic risks5 the greeks Delta 6wrt %rice of the underl#ing asset8 o Call5 ] He+"* N6dE8 o 0ut5 ]H+e+"* N6+dE8 'ho 6wrt risk+free rate8 o Call5 cHA*e+r* N6d28 o 0ut5 cH+A*e+r* N6+d28 Lega 6wrt olatilit#8 *heta 6wrt time8

>edging strategies Delta+neutral Tamma+neutral Delta+rho+neutral S$a's 4nterest rate swa%5 e&change of fi&ed+rate and floating+rate interest %a#ments for a fi&ed %ar alue o Sensiti e to interest rate risk o 0ricing swa%5 ia decom%osition of 0L6fi&ed cou%ons8 and 0L6forward rate cou%ons8 *he market %rice of the floating+rate .ond e"uals %ar after each cou%on %a#ment, Currenc# swa%5 e&change of interest %a#ments in different currencies o Sensiti e to interest rate and currenc# risks

3inancial 'isk !anagement, NES 2006/9

Lecture 4. Measurin) -olatilit&

#istorical olatilit#5 !1 !o ing 1 erage with e"ual weights E MA 6used .# 'isk!etrics8 S2t H dS2t+E J 6E+d8r2t+E H 6E+d8 Ik^0 dt+Er2t+k E&%onentiall# @eighted !o ing 1 erage "uickl# a.sor.s shocks d is chosen to minimi;e 'oot of !ean S"uared Error '!SE H b 6E/*8etHE5* 6S2t+r2t82 d H 07`G for de elo%ed markets

GARC#51617 S2t H a J . S2t+E J cO2t+E 0arameter restrictions5 a^0, .JcfE o Tuarantee that ariation is non+negati e and that unconditional e&%ectation e&ists5 ECS2D H a/6E+.+c8 !ore general model than E@!1, can .e modified o !ore lags o Le erage effect5 stronger reaction to negati e shocks !ore %arameters leads to larger estimation error o -sed less fre"uentl# in '! than E@!1

I%'lied =ased on o%tions? market %rices and 6=lack+Sholes8 model o 3orward+looking, Reali8ed =ased on intrada# data o E7g7, %rices o er hourl# inter als o $nl# for li"uid assets E0

3inancial 'isk !anagement, NES 2006/9

Lectures 9/:. Mar!et ris!2 ;aR and *e&ond

Identification of %ar!et ris! 0rimar#5 directional risks from taking a net long/short %osition in a gi en asset class o 4nterest rate / currenc# / e"uit# / commodit# Secondar#5 other o Lolatilit# / s%read / di idend o !an# trading .ooks are managed with the o./ecti e of reducing %rimar# risks2 at the e&%ense of an increase in secondar# risks Teneral 6s#stematic8 s s%ecific risks o *he latter ma# .e im%ortant .ecause of lack of di ersification or im%lementation of a s%ecific in estment strateg# Non+linear 6o%tion+like8 instruments o Need to model full %ro.a.ilit# distri.ution of the underl#ing markets factor6s8 o Sensiti e to long+term market olatilities and correlations Assess%ent of %ar!et ris!5 selection of factors and choice of models Statistical models5 factor return distri!utions o Descri.ing uncertaint# a.out the future alues of market factors o E7g7, geometric =rownian motion, T1'C> 0ricing models5 factor e&posures o 'elating the %rices and sensiti ities of instruments to underl#ing market factors o E7g7, =lack+Scholes 'isk aggregation models5 risk measures o E aluating the risk of losses in the future %ortfolio?s alue o E7g7, standard de iation, La' Traditional %easures of %ar!et ris! 0osition o Si;e and direction Lolatilit# o *ime aggregation5 b* rule o Cross aggregation5 di ersification effect E&%osure o Stocks5 .eta o 345 duration, con e&it# o Deri ati es5 delta, gamma, ega, 777

"ow are si#e and direction of the position related to the li uidit$ risk%

Issues 1ggregation and com%ara.ilit# of risks o Can?t sum u% deltas or egas o !arket s credit s o%erational risk !easuring losses Controlling risk5 %osition limits o -ntil E``0s, mostl# restrictions on si;e of net %ositions, including delta e"ui alent e&%osures #istor& of Mar!et Ris! Mana)e%ent EE

3inancial 'isk !anagement, NES 2006/9

4n late E`90s and E`a0s, o !a/or financial institutions started work on internal models to measure and aggregate risks across the institution7 o 1s firms .ecame more com%le&, it was .ecoming more difficult .ut also more im%ortant to get a iew of firm+wide risks o 3irms lacked the methodolog# to aggregate risks from su.+firm le el Earl# E``0s o Trou% of F0 re%ort7 Deri)ati)es: *ractices and *rinciples7 *heir work hel%ed sha%e the emerging field of financial risk management o $ct E``G g0 !organ %u.lished 'isk!etrics and made the data and methodolog# freel# a aila.le on the internet7 'iskmetrics was de elo%ed o er the %re ious a months, .ased on their own internal model7 E``5+2005 o Lalue at 'isk 6La'8 .ecomes a standard financial market risk measurement tool worldwide

;alue/at/Ris! 5;aR7 !a&imum loss due to market fluctuations o er a certain time %eriod with a gi en %ro.a.ilit# E+M5 0ro.6LossfLa'8HE+M [e# %arameters5 o Confidence le el5 ``B 6=asel8 or `5B 6'isk!etrics8 *he higher the confidence le el, the lower the %recision o >olding %eriod5 E0d 6=asel8 or Ed 6'isk!etrics8 "ow does 'aR *ime necessar# to close or hedge the %osition change with holding 4n estment hori;on period ( confidence Estimation %eriod inter)al% Statistical model

E2

3inancial 'isk !anagement, NES 2006/9

;aR %easure%ent2 )eneral fra%e$or! !arking+to+market %osition 0ortfolio sensiti it# to risk factors o Linear s non+linear Distri.ution of the risk factors o Normal s other 0arameter assum%tions o Confidence le el and hori;on Data o Estimation %eriod and fre"uenc# Main %ethods of co%'utin) ;aR Delta+normal o 1nal#tic, ariance+co ariance >istorical simulation o =ootstra% !onte Carlo o Simulations Stress *esting and Scenario 1nal#sis are com%lementar# tools to La' o 3ocus on %otential e&treme market mo es

Local estimation a%%roaches

Standard delta/nor%al %ethod 1ssuming that %tf returns are normall# distri.uted5 La' H L6E+e&%6kE+MStJh88 R kE+MLSt EF

3inancial 'isk !anagement, NES 2006/9

o iuantile5 E765 6`5B8 j 27FF 6``B8 o Dail# data5 assume that e&%ected return H 0 >olding %eriod u% to E0 da#s o *+da# Lar H dail# Lar X b* o 1ssuming ;ero auto+correlation 1%%lications5 o Single asset o Large, well+di ersified %tf of man# iid %ositions 6e7g7, consumer credits8 o kiuick and dirt#? wa# to com%ute La' of the .usiness unit, .ased on historical 0:L

0elta/nor%al %ethod5 risk ma%%ing Decom%osition of the %tf to multi%le risk factors5 L H Ii@i3i La' H kE+MLb@*I@ o L is decom%osed .ased on *a#lor series o @5 ector of %tf weights or sensiti ities of %tf return to factor returns o I5 co ariance matri& of risk factors Can .e decom%osed to SD and correlations5 co i,/ H corri,/ Si S/ Correlations are more sta.le in time, use larger estimation %eriod SD is more time+ ar#ing, com%uted .# E@!1 or T1'C> Decom%osition of the %tf to standardi#ed positions Ai La' H kE+MbA*IA o Standardi;ed %ositions5 sensiti e onl# to the gi en factor, with same delta as the %tf E&am%le5 for a -S in estor, %tf of dollar .onds on Z2 mln has std %ositions for interest rate risk on Z2 mln and for 3A risk on Z2 mln o *wo+stage a%%roach .# 'isk!etrics5 La' H b 0La'* l 0La' Estimate risks of each std %osition5 0La' 1ggregate using %re+estimated correlation matri& of std %ositions l Dealing with de iations from normal distri.ution5 o)erstate risk in presence of options% 3at tails / skewness o 1d/usted "uantiles .ased on Student?s t distri.ution or mi&ture of normal distri.utions Nonlinear relationshi%s5 dL H Q dS J Y m dS2 Jn dS J o dt J 777 o $%tions5 deltas are unsta.le and as#mmetric o Delta+gamma+ ega a%%ro&imation5 La' H pQp kE+MSS _ Y m 6kE+MSS82 J pnp pdSp Criti<ue iuick com%utations Decom%oses risks Strong assum%tions a.out return distri.utions Cannot handle com%licated deri ati es *he com%utations rise geometricall# with q factors Can?t aggregate olatilit# o er time with b* for large *

+ill delta-normal method under- or

3ull estimation a%%roaches

#istorical si%ulation %ethod5 non%arametric, full estimation a%%roach *ake historical time series of risk factors or assets o -suall#, at least one #ear o Longer %eriod increases the %recision unless the %rocess %ro%erties change o er time EG

3inancial 'isk !anagement, NES 2006/9

Simulate the change in alue of a gi en %tf using historical factor reali;ations o 1ctual %rice functions 6e7g7, =lack+Scholes8 o 1%%ro&imate %tf %a#off function 6.ased on %tf sensiti ities8 'e%eat simulations, %lot the em%irical distri.ution of 0:L

Modifications5 Different %ro.a.ilities for historical o.ser ations o E@!15 geometricall# decreasing %ro.a.ilities dt for lag t 60fdfE8 !ore distant e ents ha e smaller %ro.a.ilit# o >igher weight for o.ser ations from the same month 3or seasonal commodities, such as natural gas >ull+@hite, standardi;ed returns5 use 'i,tsi/Si,t in the simulations o Si,t5 historical olatilit# of factor i o si5 current olatilit# forecast for factor i 4ntra+da# returns 6e7g7, hourl# inter als8 o Can anal#;e assets with short histor# 6e7g7, after 40$8 Criti<ue Eas# and sim%le No model risk o No need to assume normal distri.ution, forecast olatilit# Correlations are em.edded 0ath+de%endent, assumes stationarit# 'e"uires long histor# o $therwise miss rare shocks Does not gi e structural knowledge Monte Carlo 5si%ulation7 %ethod !odel 6multi ariate8 factor distri.utions o Stocks5 Teometric =rownian !otion / with /um%s o 4nterest rates5 Lasicek / C4' / multifactor models Tenerate scenarios and com%ute the reali;ed 0:L o -sing factor inno ations from the model 0lot the em%irical distri.ution of 0:L Criti<ue !ost %owerful and fle&i.le o 6Cross+8factor de%endencies o !ost com%licated instruments 6%ath+de%endent o%tions8 4ntellectual and technological skills re"uired Length# com%utations !odel risk @est 6200G8, Com%arati e summar# of the methods

E5

3inancial 'isk !anagement, NES 2006/9

0ifferent t&'es of ;aR La' delta5 %artial deri ati e wrt factor or %osition La' !eta5 B measuring the contri.ution of a gi en factor or %osition to the o erall %tf risk Incremental La'5 change in La' due to a change in the %osition o 0recise measurement re"uires re+estimation Marginal ,component- La'5 DeltaX0osition o 1dditi e5 %tf La' is a sum of marginal La's Relati)e La'5 La' of the %tf?s de iation from the .enchmark o !easures e&cessi e risk 3ac!testin) ;aR Lerification of how %recisel# La' is measured o Com%are B cases when the losses e&ceed La' with the %redicted fre"uenc# "istorical a%%roach5 .ased on the actual recorded 0:L o !ore traditional, re"uired .# =asel o >el%s to identif# the model?s weaknesses, mistakes in the data, and intra+da# trading o $ften, actual 0:L %roduces lower than e&%ected fre"uenc# of La' iolations due to da# trading that allows %ositions to .e closed "uickl# when the markets .ecome olatile *hus, reducing actual losses com%ared holding a static %ortfolio for 2G hours "$pothetical a%%roach5 .ased on h#%othetical 0:L com%uted using current %tf weights 6factor e&%osures8 and historical data on assets 6risk factors8 o Concentrates on the current risk %rofile o Eliminates the im%act of intra+da# trading o =ut5 ma# gi e .iased results if the same model is used .oth for estimating 0:L and for La' Small sam%le %ro.lem o Need long histor# for high confidence le el 6``B8 to ensure statistical accurac# of La' =asel5 E #ear of dail# data
= e(ce'tions 0+G 5 6 9 a ` E0J >one Treen Wellow Wellow Wellow Wellow Wellow 'ed Scalin) factor for reser-es F F7G F75 F765 F795 F7a5 G 6model withdrawn8

3er!o$it8 and +?3rien6 @F .AA. E&amine the %ractice of La' measurement in 6 large -S .anks o Com%are La' .ased on internal model with that .ased on reduced+form model for 0:L E6

3inancial 'isk !anagement, NES 2006/9

Data, 0E/E``a _ 0F/2000 o Consolidated end+of+da# 0:L, internal dail# ``B La' o *he returns are normali;ed .# SD to hide the identities of the .anks rE5 for 5 from 6 .anks La' is e&ceeded in F or less cases from 590 rE, 3E5 when La' is e&ceed, the losses are high, o er 2 SD for F .anks 6%ro.7 07EB for t58 r2, 325 most cases of e&ceeding La' during the F months of the 'ussian default in a+E0/E``a o =anks make conser ati e estimates of La'5 the e&ce%tions are rare, .ut large and clustered in time rF5 the correlation .etween .anks? dail# 0:L is "uite low 60728, the correlation .etween .anks? dail# La' is unsta.le Should 'aR !e rG5 e aluating the accurac# of La' estimates conser)ati)e% o -nconditional co erage test5 re/ect >0 for one .ank 6.ut low %ower8 o 4nde%endence 6for i7i7d7 distri.ution of iolations85 re/ect >0 for 2 .anks o Conditional co erage test 6sum of the %re ious two85 re/ect >0 for 2 .anks Com%are with the reduced+form models5 1'!16E,E8 : T1'C>6E,E8, out+of+sam%le starting after E65 da#s o Does not account for change in %ositions and risks, cannot .e used for sensiti it# or scenario anal#sis o Em.eds s#stematic mistakes in 0:L o r5, 3G5 La' is lower, the a erage q e&ce%tions close to EB, the magnitude of losses is lower, similar test results Conclusions5 o =anks? estimates of La' are too conser ati e, do not ade"uatel# reflect risks in certain %eriodss are not .etter than those .ased on a sim%ler forecasting models 61'!1+T1'C>8 o !akes sense to use .oth models *raditional5 forward+looking, decom%oses indi idual risks *imes series5 fle&i.le and %arsimonious, ad antage in forecasting, %ro ides check for the main model

;aR i%'le%entation !easurement o Local s full estimation o 0ortfolio effects 1%%lications Lerification5 .ack testing Sensiti it#5 stress testing Limitations ;aR a''lications 0ortfolio management o !in La' with gi en e&%ected return 0osition limits o *rading s in estment o >ierarchical structure Ca%ital ade"uac# re"uirements o =asel5 reser es H FXLa'``B o *he multi%lier goes u% if La' is underestimated 'isk+ad/usted %erformance e aluation E9

+h$ does Basel accord set the )alue of multiplier to .%

3inancial 'isk !anagement, NES 2006/9

o '1'$C H 'isk+1d/usted 'eturn / Economic Ca%ital o Similar measures in cor%orate finance5 Ea' 6Earnings at 'isk8, C3a' 6Cash 3low at 'isk8 Stress testin) 1nal#sis of %tf alue under rare, .ut %ossi.le scenario o Shows the magnitude of losses which e&ceed La' o >el%s to identif# the weaknesses 3actor %ush5 assume that one of the ke# model %arameters changes a lot o E7g7, increase in olatilit# / correlations / e&change rate / oil %rice o 3or com%licated deri ati es need to check how the %tf alue changes under different alues of the %arameter o =ut5 correlations are ignored >istorical scenarios o 1ugust E``a5 'ussian g t default, ru.le de aluation, credit s%reads rising, de elo%ed countries? .onds rates falling, gradual li"uidit# crisis o Shock for a certain asset class5 =lack !onda# for the -S stock market in E`a9, war in 4ra" and oil %rice, etc7 o Shock for a certain region5 1sian crisis in E``9 6for emerging markets8, -SD weakening o `/EE terror attack, Enron re%orting scandal, Wukos case, etc7 o 1d antage5 kee% correlation .etween different factors >#%othetical scenarios o 'ising correlations5 e7g7, model each asset?s return 'i as wtd sum of 'i : market return '! o 'ussia5 .ank crisis, inflation, @*$ entr#, 2 o 4t is im%ortant to ensure that there no internal inconsistencies in the %ertur.ed model 6e7g7, ar.itrage o%%ortunities8 o 1d antage5 er# fle&i.le >#.rid method o Change the co ariance matri& : other %arameters o Estimate the conditional distri.ution of %tf alue and com%ute Bstress? ;aR E&treme+ alue theor# o Estimate the %arameters of distri.ution of ma&imum losses o =lock ma&ima5 e7g7, annual highest losses o 0eak+o er+threshold5 EB da#s with the lowest return o =ut5 needs large and re%resentati e data .ase E(a%'les of scenarios Stock market falling 6e7g7, -S in E`a98 o De elo%ed stock markets falling .# 20B, emerging ones .# F0Bs olatilit# rising .# 25B o 3light to "ualit# leads to strengthening of the dollar 6.# E0B relati e to E!8 o 4nterest rates in D! falling, E! B go u% .# 075B 6EB8 for long 6short8 maturities o Commodit# %rices falling due to fear of recession5 oil %rice goes down .# 5B *ightening of the 3ed?s monetar# %olic# 6to tighten u% inflation, as in E`aG8 o $ ernight rates rising .# EB, longer+maturit# rates .# 075B o 4nterest rates in other countries rising less o 'elati e strengthening of the dollar 6in estors %ut more in dollar+denominated instruments8 o Credit s%reads rising o Stock markets falling .# F+5B, olatilit# rising Stress testin) 'ractice Ea

3inancial 'isk !anagement, NES 2006/9

!organ Stanle#5 .lue .ook g0 !organ Chase5 L4D 6 ulnera.ilit# identification8 o 'e+estimation at least once a month, discussed .# the to% managers o Se eral macroeconomic crisis scenarios o Com%lementar# scenarios for different market segments o Conser ati e assum%tion5 %ositions remain the same during the crisis *o .e read# to the a.ru%t fall in li"uidit# $./ecti e5 .e %re%ared to an# %ossi.le financial crisis o E en though we don?t know its %ro.a.ilit# of occurring

Ris!Metrics5 g0 !organ : 'euters 6since E``G8 !arket risk measurement methodologies o Delta+normal / historical / !onte Carlo o Stress testing5 L4D 6 ulnera.ilit# identification8 Data on olatilities and correlations o Cash flow ma%%ing to stock indices, currencies, and 34 .askets Software E(a%'le2 $range Count#, ZE79 .ln losses in E``G due to rising B Le eraged %urchases of interest rate deri ati es financed with re erse re%os o =etting on the slo%e of the #ield cur e gorion5 at the end of E``G, one+#ear 5B La' was a.out ZE .ln o 1re the downside risks worth the u%side< o =ad incenti es for managers5 either modest, a.o e+market return or financial disaster !iller and 'oss5 $range Count# was not insol ent, with net assets of Z6 .ln, including Z600 mln in cash 6for Z20 .ln in total assets8 o *he .ankru%tc# could ha e easil# .een %re ented, Cui8 *otal s selecti e risk management o Lalue s cash flow at risk La' s e&%ected return and u%side %otential >ow to mani%ulate La'< @hat is the relation .etween =asel and 'isk!etrics La'< 4s it .ad if the actual losses e&ceed La'<

=e#ond La'
;aR assu%'tions 0ortfolio sensiti it# o 'isk factor co erage o 6Non+8linearit# Distri.utions o Sta.le and e&%loita.le relationshi%s o 3at tails and skewness 1r.itrar# %arameters ;aR dra$*ac!s E`

3inancial 'isk !anagement, NES 2006/9

4m%licit assum%tions o No intrada# trading o 'isks are descri.ed .# e&%osures to se eral risk factors o -suall#, second cross+deri ati es are neglected *he actual losses are unknown *he measurement error ma# .e large, es%7 for a high confidence le el !odel risk !ani%ulation5 incenti e to choose o Strategies neutral to gi en risk factors o 0ortfolios with er# unlikel# e&treme losses Not su.additi e

Pro'erties of a coherent ris! %easure D !onotonicit#5 D5E17 F D5E.7 for E1 G E. o >igher return im%lies lower risk *ranslation in ariance5 D5EHconst7 I D5E7/const o 1dding cash lowers risk .# the same amount >omogeneit#5 D5JE7 I JD5E7 for J F A o 4ncreasing the %tf?s si;e leads to %ro%ortional increase in risk Su.additi it#5 D5E1HE.7 G D5E17 H D5E.7 o 0tf risk does not e&ceed the sum of its com%onents? risks Alternati-e %ar!et ris! %easures 3ull distri.ution >igher moments o Lariance, kurtosis, etc7 0artial moments o Semi+ ariance5 risks in the falling market Downside C10!5 .eta measured on the .asis of low returns onl# o E&%ected shortfall 6coherent,8 Cash flow risk

Lecture K. Li<uidit& ris!

Mar!et li<uidit&5 the a.ilit# to o%en or close large %ositions without a strong effect on %rice 0i%ensions of %ar!et li<uidit& /ightness5 %rice de iation o $.ser ed / 'eali;ed / Effecti e s%read Dri en .# in entor# and ad erse selection costs Depth5 %otential su%%l# and demand o *rading olume o *urno er rate 6to olatilit#8 o B non+traded da#s o 1mihud 6200285 dail# ratio of a.solute return to trading olume Dail# %rice res%onse associated with ZE of trading Resilienc$5 time necessar# for %rice to reco er Immediac$5 the e&ecution time 20

3inancial 'isk !anagement, NES 2006/9

Market microstructure estimates of illi uidit$ The Glosten/#arris %odel5 Q%t H d"t J tCDt+Dt+ED J Ot *rade+.#+trade data D5 order sign o =u#er 6seller8 initiated if the %rice is a.o e 6.elow8 mid+"uote5 JE 6+E8 "5 signed trade "uantit# o 0ositi e for .u#er+initiated trades t5 fi&ed cost com%onent d5 in erse market de%th %arameter The #as*rouc!/Foster/;is$anathan %odel5 Q%t H M J d"-t J tCDt+Dt+ED J Ot "-5 the une&%ected signed trading olume o =ased on the model for " with fi e lags of " and fi e lags of Q% o !easures the informati eness of trades Cost com%onents, in B of the %rice o 0ro%ortional5 d X a g trade si;e / a g closing %rice o 3i&ed5 t / a g closing %rice Li<uidit& and asset 'ricin) Lower transaction costs lead to lower e"uit# %remium for risk =rennan 6g3E, E``685 o 0ortfolios with higher estimated costs ha e higher e&%ected return -ne&%lained .# 3ama+3rench model o >igher s%read im%lies lower returns, S%read is a .ad measure of illi"uidit# 1mihud 6200285 o E&%ected illi"uidit# im%lies higher e&%ected return o -ne&%ected decrease in li"uidit# decreases the current %rices 0eter%inants of the li<uidit& 1sset characteristics o Su.stitution 6leading to concentration8 s com%lementarit# effects !arket microstructure "ow does li uidit$ change with o *ransaction costs the degree of 1specialness2% o 4nfo trans%arenc# o *rading s#stem 1uction s dealershi% markets5 %rice+ olume trade+off =eha ioral factor o Dominating market %artici%ants +hich market is more li uid0 o >eterogeneit# with short or long-term =u#/sell, risk attitude, in estment hori;on S'ecifics of the Russian stoc! %ar!et Li"uidit# is concentrated in .lue chi%s Li"uidit# mo ed from '*S to !4CEA 3light to li"uidit# 6or "ualit#8 during the crises o E7g7, 1ugust E``a, 1ugust 200F, 1%ril 200G 2E

3inancial 'isk !anagement, NES 2006/9

Modelin) %ar!et li<uidit& ris! *he actual %rice ma# differ from the current market %rice *he effecti e s%read de%ends on the transaction si;e and timing o >arder to measure for the $*C market o 4ncreases during the crisis !onte Carlo anal#sis5 La' ad/ustment, accounting for o Direction and si;e of %ositions No more %ositi e homogeneit# of degree E 4deall#, should know elasticit# of %rice wrt olume o Correlation .etween market d#namics and li"uidit# 1s#mmetr# .etween .ullish and .earish markets Stress testing, accounting for o !argin re"uirements Es%7 if #ou hold a large %ortfolio with one .roker o 'isk limits Low limits will soon re"uire further sales to sto% losses o *he likelihood of s#stemic crisis o *#%ical scenario5 a dealer sto%s to %ro ide "uotes Fundin) li<uidit& 5insol-enc&7 ris! 4na.ilit# to fulfill the o.ligations .ecause of the shortage of li"uid funds Determinants5 o Current cash reser es o 1.ilit# to .orrow or generate C3s Sources5 o S#stemic E7g7, 'ussian g t default in 1ugust E9, E``a o 4ndi idual Change of a com%an#?s 6im%licit8 credit rating o *echnical -n.alanced forward %a#ment structure -ncertaint# a.out future C3s E(a%'le2 LTCM 4n estment strateg# o =etting on con ergence of s%reads E7g7, long %osition on the off+the+run *reasur# .onds, short %osition in the on+the+run 6recentl# issued and more li"uid8 *reasuries 4n general, long 6short8 %osition in riskier 6less risk#8 instruments o >igh le erage, no di ersification o =rought net returns o er G0B in E``5 and E``6 o =ut5 sensiti e to market+wide li"uidit# 1t the end of E``95 o 1fter returning Z279 .ln to their in estors o =alance sheet assets of ZE25 .ln o $ff .alance sheet notional amounts round ZE trln !ostl# netta.le 6swa%s8 22

3inancial 'isk !anagement, NES 2006/9

o 0ositions with nominal alue of ZE725 trln 1ddressing funding li"uidit# risk5 o $wn ca%ital of ZG7a .ln F0 times le erage for =S sheet assets onl#, o Credit line of Z`00 mln o 4n estors commit ca%ital for at least 2 #ears *he crisis and its resolution o 1ugust E``a5 'ussian default triggered (flight to safet#), all risk %remiums rose L*C! lost Z550 mln in 1ugust 2E onl# o =# the end of Se%tem.er, ca%ital declined to ZG00 mln L*C! was forced to li"uidate %ositions to meet margin calls o Se%tem.er E``a5 `0B of the L*C! %tf was %urchased .# a consortium of EG .anks for ZF7625 .ln *hat %re ented the danger that the default of L*C! would trigger man# cross+defaults o gul# E```5 redem%tion of the fund *he issues raised o 'isk management at L*C! 'ole of stress testing o 'isk management at L*C! counter%arties 4nteraction with a highl# le eraged institution o Su%er ision o !oral ha;ard

Lectures L/14. Credit ris!

Credit ris!5 losses due to the counter%art#?s failure to honour his o.ligations Credit e-ent Default5 an o.ligation is not honored 0a#ment default5 an o.ligor does not make a %a#ment when it is due o 'e%udiation5 refusal to acce%t claim as alid o !oratorium5 declaration to sto% all %a#ments for some %eriod of time -suall#, .# so ereigns o Credit default5 %a#ment default on .orrowed mone# 6loans and .onds8 4nsol enc#5 ina.ilit# to %a# 6e en tem%orar#8 =ankru%tc#5 the start of a formal legal %rocedure to ensure fair treatment of all creditors S'ecifics of CR5 1s#mmetr#5 %ossi.ilit# of .ig losses o 1the most $ou can lose is e)er$thing2 'are occurrence Longer hori;on Non+trada.ilit# of most loans o >ard to measure correlations Limits at the transaction le el 4nteraction with market risk Es% im%ortant for .anks 2F

3inancial 'isk !anagement, NES 2006/9

Larger reser es o =asel5 GXLa'

E(a%'les Sa ings:loans crisis in the -S in E`a0s o *he restructuring cost the g t ZF0 .ln Defaults on Latin 1merican countries? de.t in E`a0s o 'estructuring in =rad# .onds Defaults on cor%orate .onds o Es%7 /unk .onds at the end of E``0s 1ccumulation of .ad "ualit# loans in ga%anese .anks 'ussia o No default on cor%orate .onds so far o Wukos on the .rink of .ankru%tc# Measurin) credit ris! =asic com%onents5 o 0ro.a.ilit# of default 60D8 o 'eco er# rate 6''8 / loss gi en default 6LTD8, '' J LTD H E o Credit e&%osure 6CE8 / E&%osure at default 6E1D8, CE H E1D Credit loss5 CLi H DiXLTDiXE1Di o Di5 default indicator *he credit loss distri.ution5 CL H IiCDiXLTDiXE1DiD o >ighl# skewed5 limited u%side, high downside o Correlation risk5 assuming inde%endence will underestimate risks o Concentration risk5 sensiti it# to the largest loans Credit ;aR H @CLM _ ECCLD o Difference .etween e&%ected losses and certain "uantile of losses o E&%ected losses co ered .# %tf?s earnings o -ne&%ected losses co ered .# ca%ital reser es Modelin) credit ris! Internal a%%roach5 usuall# used .# .anks, focus on 0D o Classical sol enc# anal#sis o !arket en ironment "ow do securiti#ation and credit o iualit# of the com%an#?s management deri)ati)es influence modeling o Credit histor# o Credit %roduct characteristics E&ternal a%%roach5 using market data on stocks / .onds !L6C'8H f6loss distri.ution, risk %remium8 o Need to measure .oth 0D and '' o 'e"uires li"uid secondar# market Reco-er& rate >ighl# aria.le, neglected .# research for a long time o 3ragmented and unrelia.le data !arket alue reco er#5 2G

3inancial 'isk !anagement, NES 2006/9

o !L %er unit of legal claim amount, short time 6E/Fm8 after the default Settlement alue reco er#5 o Lalue of the default settlement %er unit of legal claim amount, discounted .ack to the default date and after su.tracting legal and administrati e costs Legal en ironment factors o Collateral or guarantees o 0riorit# class5 collaterali;ed, senior, /unior, etc7 o *he .ankru%tc# legislation Large cross+countr# differences5 e7g7, -S and 3rance more o.ligor+friendl# than -[ -S .ankru%tc# %rocedures5 ch7 EE 6aim to restructure the o.ligor8 s ch7 9 6aim to li"uidate the o.ligor and %a# off de.t8 $ther em%iricall# o.ser ed factors o 4ndustr# o *he o.ligor?s rating %rior to default o =usiness c#cle : a erage rating in the industr# 1ltman5 -S, E`9E+E``5 o $n a erage, a.out G0B with SD of 20+F0B !odelling ''5 .eta distri.ution with densit# f6&8 H c &a 6E+&8. o 3or mean h and ariance S25 aHh26E+h8/S2, .Hh6E+h82/S2

Credit e('osure5 the amount we would lose in case of default with ;ero reco er# $nl# the %ositi e economic alue counts5 current CEtHma&6Lt, 08 o Lt5 credit %ortfolio?s alue +hat is 3E for futures Current s %otential or short option% o E&%ected 6ECE8 s @orst 6@SE8, for a gi en confidence le el Loans and .onds5 direct, fi&ed e&%osures o -suall#, at %ar Commitments 6e7g7, line of credit85 large %otential e&%osure o -suall#, fraction of %ar $*C deri ati es5 aria.le e&%osures o -suall#, at current alue with ad/ustment for market risk E7g7, `0B "uantile of the distri.ution o 4nterest rate swa%5 Diffusion effect5 increasing risk o er time 4morti#ation effect5 decreasing duration

*raditional a%%roaches to C' measurement

Credit ratin)s 4nde%endent agencies5 S:0, !ood#?s, 3itch 4ntegral estimate of the com%an#?s sol enc# .ased on 0D 6and ''8 o S:05 (general creditworthiness2 .ased on rele ant risk factors) o !ood#?s5 (future a.ilit#2 of an issuer to make timel# %a#ments of %rinci%al and interest on a s%ecific fi&ed+income securit#) *he rating %rocedure o Com%rehensi e anal#sis, from micro+ to macro+le el iuantitati e5 .ased on financial re%orts iualitati e5 e aluation of the management 6.usiness %ers%ecti es, risk attitude, etc78 25

3inancial 'isk !anagement, NES 2006/9

Legal o *he rating committee decides .# oting o *he rating is reconsidered at least once a #ear Long+term com%an# rating o 4n estment rating5 from === 6S:08, =aa 6!?s8 !eant for conser ati e in estors o S%eculati e rating o Smaller gradations5 E/2/F, J/+, $utlook, @atch (/hrough-the-c$cle) a%%roach5 C' at the worst %oint in c#cle o er contract?s maturit# o Does not de%end on the current market en ironment o 4n contrast to the (point-in-time) a%%roach5 C' de%ending on the current macro en ironment Short+term com%an# rating 'ating of s%ecific instruments

1%%lications5 Credit %olic# and limits 0ricing the credit !onitoring and credit control Securiti;ation Sur-i-al anal&sis5 actuarial a%%roach to estimating 0D Sam%le %eriod from 20 #ears o Need to accumulate default statistics 6es%7 for first+class issuers8 o Need to account for different stages of the .usiness c#cles 3or each rating5 a erage 0D o E"ual weights5 0D H q defaults / q com%anies 6with a gi en rating8 1nother a%%roach5 weights %ro%ortional to the issue?s olume *#%e of the .onds o =onds traded in the -S market s .onds of -S com%anies o Straight .onds s con erti.le/redeema.le .onds o Newl# issued .onds s seasoned .onds >ori;on o Lower q o.ser ations for longer hori;on !easures of 0D !arginal mortalit# rate 6!!'8 in #ear t o Estimated 0D in #ear t after the issuance Sur i al rate 6S'8 o 4n #ear E5 S'E H E+!!'E o During * #ears5 S'* H 6E+!!'E8X2X 6E+!!'*8 !' in #ear t gi en sur i al during t+E #ears5 !'t H S't+EX!!'t Cumulati e !' during * #ears5 C!'* H ItHE5*!'t H E+S'* 1 erage !' during * #ears5 1!'* H E+6E+C!'*8E/* Estimation results Cumulati e 0D goes down with rating *he highest !!' is for G+5 #ears since the .ond?s issuance

26

3inancial 'isk !anagement, NES 2006/9

Criti<ue of credit ratin)s 'atings react with a lag to the change in the com%an#?s sol enc# =ias in ratings5 too conser ati e Large measurement error o Com%anies with the same 6different8 rating ma# ha e different 6same8 rating o 'atings of different agencies often do not coincide !onte Carlo anal#sis .# [!L o Tenerate 50,000 times a 25 #ear sam%le of N com%anies with a gi en 0D *he %arameters match those in !ood#?s data .ase o *he estimated 0D is er# nois#, higher than the actual one for most issuers o *he differences .etween the estimated and actual 0D rise with correlation .etween defaults Credit ratin)s -s credit s'reads Credit s%read of a gi en .ond often differs from the a g in the grou% with the same credit rating Credit s%read de%ends on C' and other factors5 o Li"uidit# risk o !aturit# o !acroeconomic factors o !arket olatilit# 'atings react with a lag to the change in credit s%read o *hough the difference usuall# disa%%ears with time5 change in the firm?s sol enc# is reflected in rating within half a #ear, or s%read returns to the initial alue Internal ratin) s&ste%s5 e aluation of the .orrower?s sol enc# Criterions o 0ro.a.ilit# of default o 'eco er# rate >ori;on o -suall#, E #ear Scale o 1ccording to S:0/!ood#?s or its own 3actors o E7g7, 5C5 Character, Ca%ital, Ca%acit#, Collateral, C#cle Anal&sis of financial re'orts Current / "uick li"uidit#, le erage, %rofita.ilit#, turno er ratios =ackward+looking5 using historical data o E&tra%olation of the %ast into the future gi es im%recise forecast Es%7 under high uncertain# 4n 'ussia5 little trust to fin re%orts o '1S is clearl# inferior to 41S o $ften re%orts are corru%ted *o misguide ta& authorities, minorit# shareholders, .anks, etc7 'ole of e&%ert?s o%inion Lisit to the com%an# 0ersonal contact with o%t managers >uman factor 29

3inancial 'isk !anagement, NES 2006/9

Credit scorin) %odels5 %redicting the default .ased on the .orrower?s data 1ltman?s >/score 6E`6a85 linear function of 5 aria.les o AE5 @orking Ca%ital to 1ssets o A25 'etained Earnings to 1ssets o AF5 E=4* to 1ssets o AG5 !arket Lalue of E"uit# to =ook Lalue of Lia.ilities o A55 Sales to 1ssets *he sam%le5 66 com%anies, half of which defaulted \ H E72AE J E7GA2 J F7FAF J 076AGJ 07```A5 4nter%retation5 o \ ^ F5 default is unlikel# o 279 f \ V F5 closer to the (dangerous ;one) o E7a f \ V 2795 likel# default Is t$pe 5 or t$pe 6 classification o \ V E7a5 er# high 0D error more important% >ow to e aluate the model< o *#%e+E error5 default .# the .orrower who recei ed a loan o *#%e+2 error5 %redicting default for the good .orrower o 'esults5 more than `0B firms were correctl# classified Constructin) a scorin) %odel 3inancial aria.les5 o 0rofita.ilit#, income olatilit# o Le erage and interest co erage o Li"uidit# o Ca%itali;ation o !anagement "ualit# !ethodolog# o *he .inar# choice logit / %ro.it model o Discriminant anal#sis E&am%les5 o >ETA+model 6E`9985 for .ig com%anies 6ZE00 mln in assets8 9 aria.les instead of 5 o EMS 6emerging markets score85 for emerging countries -sing the model cali.rated .# the -S firms 1d/usted for the risk of currenc#?s de aluation, industr# s%ecifics, com%etiti e ad antage, %resence of a guarantee, etc7 Criti<iue5 %ros and cons Sim%licit# Sol es the %ro.lem of su./ecti it# of e&%erts? grades -suall# assumes sim%le linear de%endence Limited theor# to e&%lain the degree of each aria.le?s im%act o Danger of o erfitting in+sam%le Need a good data+.ase o -suall#, a .iased sam%le e&cluding .orrowers that were denied credit Same old %ro.lems with financial re%ort data

2a

3inancial 'isk !anagement, NES 2006/9

>ow to estimate losses of a credit %ortfolio<

Re<uire%ents to an ideal CR %odel 4nstrument characteristics o Seniorit# o Collateral / guarantee Com%an# characteristics o 3inancial le erage o =alance li"uidit# o *#%e of the .usiness5 c#clical, growing o Si;e 4m%act of macro factors o 'ecession o Decline in the stock or commodit# market o !arket li"uidit# crisis 0ortfolio effects o 4nteraction .etween different credit instruments o Sensiti it# to common risk factors 3asel re<uire%ents *he internal C' models should ca%ture o Concentration / s%read / downgrade / default risk Ca%ital charge H GXLa'6``B, E0d8 4nteraction .etween !' and C' o S%read risk includes .oth interest rate risk 6!'8 and default %remium 6C'8 o Credit e ents are often antici%ated and reflected in s%read o Default is a s%ecial case of downgrade 4deall#5 integrated a%%roach to !' and C' measurement Classification of CR %odels 1ggregation o =ottom u%5 start from indi idual %ositions o *o% down5 model aggregate credit %ortfolio *#%e of C' o Default+mode 6loss+.ased85 an e&%osure is held till maturit#, when it is re%aid or defaults o !ark+to+market 6N0L+.ased85 track changes in the current market alue of the instrument !ethod of estimating 0D5 o Conditional5 de%ending on the current stage of the .usiness c#cle o -nconditional !odeling default o Structural5 6endogenous8 decision .# the firm o 'educed+form5 6e&ogenous8 stochastic %rocess

2`

3inancial 'isk !anagement, NES 2006/9

0ro%rietar# models of credit risk

3ench%ar! CR %odels Credit!etrics, E``9 o g0 !organ Chase [!L 0ortfolio !anager, E``a o [ealhofer, !ciuown, Lasicek Credit'iskJ, E``9 o Credit Suisse Credit 0ortfolio Liew, E``9 o !c[inse# CreditMetrics 1 Credit ;aR I5 C' dri en .# change in credit rating C' is modeled as change in credit rating during a %eriod e"ual to La' hori;on o Each rating is characteri;ed .# a certain 0D 1ssuming that all com%anies with same rating ha e same risk of default o -se migration %ro.a.ilities 6of mo ing from one to another rating8 o -se data from one of the ratings agencies or internal ratings Com%ute a %ro.a.ilit# distri.ution of the future alue of the instrument, using o 3orward rates for each rating *hus, account for duration and s%read effects o 'eco er# rate in case of default De%ending on seniorit# Estimate La' as a difference .etween a gi en "uantile and e&%ected alue =ottom u% a%%roach5 o 3irst estimate La' for each instrument o *hen aggregate to %tf La' accounting for correlations

!igration of credit ratings *ransition matri&, usuall# for E #ear o 0ro.a.ilit# that a com%an# with rating A ne&t #ear recei es rating W >ow to estimate *+#ear transition matri&< o Directl# =ut5 increasing * means fewer o.ser ations and larger measurement error o Cross+multi%l#ing annual transition matrices * times =ut5 ignoring auto+correlation effects

F0

3inancial 'isk !anagement, NES 2006/9

E(a%'le5 =ond with === rating, senior, unsecured, maturing in 5 #ears, 6B cou%on rate 'atings5 S:0 o 9 grou%s5 from 111 6first+class .orrowers8 to CCC 6default8 >ori;on5 E #ear o Could .e from E to E0 #ears 1nnual forward cur e for each rating o 1llows us to com%ute the alue of an# .ond in E #ear from now 0rice of the .ond in E #ear, if it kee%s === rating

3orward distri.ution of the .ond?s %rice

FE

3inancial 'isk !anagement, NES 2006/9

070E "uantile of QL distri.ution5 +2F7`EHLa'``B Credit!etrics for a %ortfolio5 di ersification effect 1ssume that return on assets is distri.uted as stock return 6T=!8 Determine thresholds for the return distri.ution corres%onding to the actual migration %ro.a.ilities Deri e the correlation matri& o =ased on multifactor model with user+defined countr# and industr# weights !onte Carlo anal#sis o Tenerate /oint rating migration scenarios for .onds within the %ortfolio o Estimate the em%irical distri.ution of %tf alue and com%ute La' Criti"ue 4gnore !' o Deri ati es re"uire stochastic interest rates o Es%7 im%ortant to assume stochastic interest rates for deri ati es 1ssume homogeneit# with the same rating class Discrete migration matri& .ased on a g historical fre"uencies o *ransition %ro.a.ilities are usuall# underestimated Sim%lified estimation %rocedure for correlations

*heoretical a%%roaches to C' measurement

Structural a%%roach5 estimate risk+neutral 0D o -se risk# de.t %rices o -se e"uit# %rices and !erton 6E`9G8 o%tion model Reduced-form a%%roach5 use actuarial estimates of 0D o 1%%l# on the %ortfolio le el

Credit s'read anal&sis 6.ased on .ond?s market %rice8 3or a one+%eriod ;ero+cou%on .ond5 r+rf R0DXLTD o LTD5 relati e to face alue Com%onents5 F2

3inancial 'isk !anagement, NES 2006/9

o Credit risk %remium -suall# rising with maturit# o Li"uidit# risk %remium Determinants5 o !acro factors !arket olatilit# / li"uidit# o =ond characteristics

Merton %odel5 stock as a call o%tion on the alue of the com%an# L

1ssume that L follows a log+normal distri.ution5 o \tuN60,E85 std =rownian motion o SL5 olatilit# 6SD8 of the relati e growth in L 6dL/Lt8 o h5 a erage growth rate, ECLtDHL0eht *he com%an#?s ca%ital structure includes o E"uit#, with alue E o De.t 6;ero+cou%on8, with face alue 3 and maturit# * Default occurs at maturit# if L*f3 o Stockholders recei e at *5 ma&6L*+3,08 o Creditors recei e at *5 min6L*, 38 Stockholders5 call o%tion on the alue of the com%an# L o E&ercise date5 * o 0rice of the underl#ing asset5 L o Lolatilit#5 SL Creditors5 .ond and short %ut o $r .asic asset 6the com%an#8 and short call

Deri ation of the %arameters5 L and SL are uno.ser a.le, deri ed from two e"uations5 o *he alue of e"uit# .# =lack+Scholes5 EHLXN6dE8+3e+r*N6d28 r5 risk+free rate dEHCln6L/38 J *6rJS2/28D / CSLb*D, d2HdE+SLb* o *he e"uation for stock olatilit#5 SEE H N6dE8 SLL 'isk+neutral %ro.a.ilit# of default5 0D H E+N6d28 H N6+d28 'eco er# rate 6as B of the assets85 '' H CE+N6dE8D / CE+N6d28D o *he current market alue of de.t5 L+EHe+r*C6E+0D8J0DX''DX3 4m%licit assum%tions5 Stockholders? .eha ior _ as gi en o *hough the# are interested in raising risk Lognormal distri.ution o -nderestimate 0D at short hori;on =ankru%tc# when L .elow the face alue of de.t o Default ma# .e different from .ankru%tc#

FF

3inancial 'isk !anagement, NES 2006/9

KM;5 main idea Estimate 0D using the modified !erton model !o e from em%irical 0D to risk+neutral 0D Deri e anal#ticall# the future alue of the com%an#?s o.ligations and La' KM; Credit Monitor 6E``F85 ED3 model Estimate the market alue and olatilit# of the firm?s assets using modified !erton model o 3or %u.lic com%anies5 estimate L and SL .ased on e"uit# %rices 1ssume more com%licated ca%ital structure5 e"uit#, short+term de.t, long+term de.t, and con erti.le %referred shares o 3or %ri ate com%anies5 estimate L and SL .ased on financial accounting measures L is .etween the o%erational alue 6%ro%ortional to E=4*8 and .ook alue SL is an em%irical function of sales, assets, industr#, etc7 Com%ute distance to default 6measure of default risk in * #ears8 o Em%iricall# estimated default %oint5 D0 H short+term lia.ilities J Y long+term lia.ilities o Distance to default5 DD H lnCE6L8+D0D/SL 6in B, in S8

!a%%ing DD to actual 0D using historical data, for a gi en time hori;on o E('ected 0efault Fre<uenc&5 ED3 H q defaulted com%anies / q com%anies, for gi en DD o Can com%ute im%lied rating

FG

3inancial 'isk !anagement, NES 2006/9

Criti"ue Com%an#+s%ecific Continuous, not .iased .# %eriods of high or low defaults *he correlation .etween defaults .ased on stock %rice correlations *esting5 ED3 rises shar%l# E+2 #ears .efore the default o 1gencies? ratings are slow to react =est a%%lied to %u.licl# traded firms o Can?t estimate countr# risk 4gnore more com%licated features of the de.t o Seniorit#, collateral, etc7

F5

3inancial 'isk !anagement, NES 2006/9

KM; Portfolio Mana)er5 C' dri en .# change in !L6assets8 Estimate actual ED35 [!L Credit !onitor Deri e risk+neutral ED3 o Cumulati e risk+neutral ED3 at hori;on *5 i* H NCN+E6ED3*8JShar%eb*8D o Su.stitute Shar%eiHciXShar%e!X*v 4n theor#, vHY o Cali.rate the market Shar%e ratio and v with o.ser ed cor%orate s%reads o er L4=$' 3or maturit# t5 r+rfH6+E/t8ln6E+iXLTD8 Estimate stock return correlation matri& using a F+le el multifactor model o 4ndi idual o Countr# and industr# o Tlo.al and regional 0ortfolio?s losses5 LHL*6NoDefault8+L*6e"uili.rium8 o 1nal#tical deri ation of La' o *he limiting loss distri.ution5 normal in erse 6highl# skewed and le%tokurtic8 Criti"ue *heoreticall# sound a%%roach o -sing risk+neutral %ro.a.ilities ED3 is a good measure of default risk Need market %rices of e"uit# o 1ssuming li"uid market o Can?t estimate countr# risk >ard to account for different t#%es of de.t =eha iour of e"uit#holders _ as gi en o *hough the ha e incenti es and are a.le to increase risk 1ssuming lognormal distri.ution CreditRis!H5 actuarial a%%roach to estimate 0D *ime hori;on5 usuall#, E# 1ssume o 3or each loan, 0D is small and inde%endent across %eriods o No assum%tion a.out the causes of the default P0 for a %tf5 0oisson distri.ution, 06n defaults8 H hne&%6+n8/n, o 1 g q defaults5 hHIi0Di o St7 de 75 bh F6

+h$ is actual st7 de)iation of *D larger than 89%

3inancial 'isk !anagement, NES 2006/9

1ssume stochastical mean 0D5 h is gamma+distri.uted o $therwise, olatilit# of 0D is underestimated E('osure H 3orward alue X LTD o Differing e&%osure amounts ma# result in a loss distri.ution far from 0oisson *he loan %tf is di ided into e&%osure .ands o Each .and / has same e&%osure w/ 6in rounded units8 o 3or each .and, e&%ected loss5 O/Hw/h/s Deri ing anal#tical distri.ution of %tf losses5 o 0ro.a.ilit# generating function for each .and o 0ro.a.ilit# generating function for the entire %tf o Loss distri.ution of the entire %tf De%ends on two sets of %arameters5 O/ and w/ E&tensions5 o >old+to+maturit# hori;on Decom%ose the e&%osure %rofile o er time o !ulti%le #ears *ake into account that default ha%%ens onl# once o Sector anal#sis5 dealing with concentration risk 1ssign sector+s%ecific %arameters to gi en o.ligors o Scenario anal#sis Criti"ue o Eas# im%lementation 3ocus on default 3ew in%uts o 1nal#tical form of the results o 4gnore !' and migration risk o 'educed+form o Not a%%lica.le to non+linear instruments

Credit Portfolio ;ie$5 to% down a%%roach using macro factors 1ssume 0D6grade8 H logistic f6macroeconomic inde&8 o *he inde& H linear f6macro and industr# factors8 o Each factor follows 1'628 B, 3A, industr# growth Estimate conditional rating migration matri& o 1d/ust the unconditional migration matri& .# the ratio of conditional and unconditional simulated 0D o 'ecession5 more mass to downgrade migrations and 0D !onte Carlo anal#sis o Simulate the /oint cond distri.ution of default and migration %ro.a.ilities Criti"ue o Link macro factors to default and migration %ro.a.ilities o Need relia.le historical data on 0D o =est a%%lied to s%eculati e grade o.ligors o 1d hoc %rocedure of estimating the migration matri& +ther %odels CreditTrades5 'isk!etrics, Toldman Sachs, g0 !organ, Deutche =ank F9

3inancial 'isk !anagement, NES 2006/9

o Stochastic default .arrier5 lognormall# distri.uted, with %ossi.le discrete /um%s 4ncreases estimated 0D o 0D is a closed form function of 6 %arameters5 !ean and std of '' 4nitial and current stock %rice 4m%lied stock olatilit#5 cali.rated from actual CDS s%reads o CreditTrade H model im%lied 5# credit s%read 1lgorithmics !ark+to+3uture 6!t385 scenario+.ased a%%roach o Links C', !', and L' o Tenerates cumulati e 0D conditional on scenario

Credit risk management

I%'act of CR on deri-ati-es Deri ati es on 6credit8 risk# .onds o Long %ut gains in alue Lulnera.le deri ati es5 su./ect to C' .# the writer o *he alue diminishes .# the credit s%read5 L?HLX0risk#6*8/0'36*8 Traditional %ethods of CR %ana)e%ent5 credit e&%osure s default modifiers !odeling CE o =4S a%%roach5 CE H !L6deal8 J %otential CE -suall#, %otential CE as 0+E5B of %ar o Direct estimation of the %otential CE distri.ution5 Stress+case / trees / !onte Carlo Collateral arrangements5 .ecome %o%ular for 6longer+dated8 swa%s o Deri ati es5 usuall#, initial marginH0, ariation margin onl# in case of !t! mo es o er the e&%osure limit o Longer %eriod .etween remarkings o 0ost securities as collateral 'emarking / recou%oning 6to .ring the alue .ack to 08 o -suall# "uarterl# Netting5 reduce %otential CE to net %osition o =etter assessed .# scenario and simulation anal#sis o =ut5 risk of (cherr#+%icking) Credit guarantees o Es% alua.le if the guarantee is from an inde%endent com%an# o -suall# %ro ided .# %arent com%anies Credit triggers o 'ating downgrade clause5 right to terminate all transactions if the counter%art#?s rating falls .elow the trigger le el !utual termination o%tion o *ime %ut5 right to terminate on one or more dates using a %re+agreed formula to alue the transaction at these times o *ermination .rings CE to 0 !odeling reco er# rate o Deri ati es5 usuall#, rank e"uall# with senior unsecured de.t

Fa

3inancial 'isk !anagement, NES 2006/9

4m%lementation issues Legal considerations o Collateral5 ma#.e %ro.lems with enforcea.ilit# in case of .ankru%tc# Economic considerations o 'esources re"uired to im%lement recou%oning and collateral arrangements CR %ana)e%ent instru%ents Limits o !aturit# / collateral / currenc# / regional / industr# *arget le els o 'eturn to C' Di ersification 'eser es Credit deri ati es

Credit deri ati es

E-olution of deri-ati-es5 F wa es Second and third generation of %rice and e ent deri ati es o >#.rid / contingent / %ath+de%endent risks Strategic management o 0tf risk, .alance sheet growth, o erall .usiness %erformance New underl#ing risks o Catastro%he / electricit# / inflation / credit h& credit deri-ati-es" Se%arate C' mgt from the underl#ing asset o Confidentialit# o Customi;ed terms o $./ecti e and isi.le market %ricing Short+selling C' .ecomes %ossi.le o 1r.itrage and efficient markets $ff+.alance+sheet o%erations o =anks can a oid selling loans *a& considerations / under%ricing / client o >edge funds can in est in C', with le erage 1 oid transfer of %ro%ert# rights and administrati e costs Com%leting the market o 'isk managers hedge C' o 4ssuers minimi;e li"uidit# costs o 4n estors find interesting instruments 'a%id growth since end of E``0s, currentl# o er Z2 trln *#%es of insured C'5 o Credit e ent, alue of the underl#ing asset, reco er# rate, maturit# Classified .#5 o *#%e of the underl#ing asset o *rigger e ent o 0a#off function F`

3inancial 'isk !anagement, NES 2006/9

Credit default s$a'5 'egular %remium %a#ments in e&change for a one+time %remium in case of the credit e ent o !aterialit# clause5 credit e ent is not triggered .# a technical default o =oth credit e ent and %a#ments can .e linked to a grou% of o.ligations Settlement o 3i&ed %a#ment o Cash settlement5 difference .etween the strike 6%ar8 and current market %rice o 0h#sical settlement in return of the %ar amount =asket default swa% o *he underl#ing asset5 loan %ortfolio 3irst+to+default 6.asket8 swa% D#namic credit swa% o Changing %rinci%al 0ractical role o Enhance li"uidit# o >edging / in estment o%%ortunit# Total return s$a'5 3i&ed or floating %a#ments in e&change for the current income from the underl#ing asset o 'egular e&change of %a#ments o 4nsures .oth !' and C' Credit o'tions5 Call or %ut on the %rice of 3'N, .ond, loan, or asset swa% %ackage o 6!ulti+8Euro%ean / 1merican +hat is a !etter hedge of 3R: o Can knock out u%on credit e ent credit option on floating 0ractical role ,:R;- or on fi&ed-rate note% o Wield enhancement o Credit s%read %rotection o >edging future .orrowing costs Downgrade o%tions +ther credit deri-ati-es >#.rid o 'e"uire a material mo ement in B, e"uit# %rices, 2 Credit s%read forwards / o%tions Credit+linked note5 cou%on %a#ments conditional on the credit e ent o -suall# ia S0L 6s%ecial %ur%ose ehicle8, trust com%an# Ris!s of credit deri-ati-es Correlation5 simultaneous default of the underl#ing asset and %rotection seller =asis Legal o E```, 200F5 4SD1 ado%ted standard terms and documentation Li"uidit#5 usuall# traded $*C 0rotection5 o =ilateral netting G0

3inancial 'isk !anagement, NES 2006/9

o $%tion for %remature a.ortion of the contract in the case of the counter%art#?s financial distress

Lecture 19. +'erational ris!

+'erational ris!5 Definition E5 financial risks .esides !' and C' o =ut5 includes .usiness risks o Did the credit default result from the knormal? credit risk or mistake of the loan officer< Definition 25 risks originating at financial transactions o =ut5 e&cludes risks due to internal conflicts, model risk, 2 Definition F5 risks due to deficiencies or mistakes from o 4nformation s#stems and technologies o 4nternal %rocedures o 0ersonnel o E&ternal e ents Classification of +R $%erational failure 6internal8 risk o 0eo%le5 incom%etenc# / fraud o 0rocess5 model / transaction / o%erating control risk o *echnolog#5 info s#stems, software, data .ases $%erational strategic 6e&ternal8 risk o En ironmental factors o Change in %olitical and regulator# regime -suall#, .usiness risk is e&cluded o Choice of strateg# o Loss of re%utation o Legal E(a%'les of +R failures =arings and Nick Leeson 6the (rogue trader)85 more than ZE .ln losses in E``5 o Strateg#5 cash+futures ar.itrage, Singa%ore+$saka ar.itrage o =ooked losing trades to account aaaaa o End of E``25 hidden losses of 2 mln %ounds o End of E``G5 hidden losses of 20a mln %ounds o E``G5 unauthori;ed %ositions in o%tions o ganuar# E``55 earth"uake in [o.e, massi e margin calls Daiwa .ank and 4guchi *oshihide o Since E`9`, L0 in NW, a erage %rofit of ZGmln o E`aG5 loss of Z50,000+200,000, o E``65 losses accumulated to ZE7E .ln, confessed o *he .ank hid this from 3ed, was fined ZFG0 mln and had to close its -S o%erations Sumitomo cor%oration and >amanaka Wasuo, Z276 .ln total losses o Since E`95, in co%%er section, at he#da# nicknamed (!r7 5B), (*he >ammer) o 3rom E`aG5 unauthori;ed s%eculati e futures trading together with the head of the co%%er trading team, tr#ing to .oost %rofita.ilit# o E`a95 cumulati e losses Z5amln, >amanaka .ecame head of the co%%er section, recei ed ZE50 mln from !errill L#nch GE

3inancial 'isk !anagement, NES 2006/9

o E``05 .egan .orrowing mone# against Sumitomo?s trading stocks to fund his trading %ositions, started fictitious o%tion trades to create an im%ression of success o E``E5 asked a .roker to issue a .ackdated in oice for fictitious trades, worth ZF50 mln *he e&change notified Sumitomo, which re%lied it was needed for ta& reasons o E``F5 .orrowed ZE00 mln from 4NT using forged signatures of senior managers o E``G5 raised ZE50 mln from !organ, then ZF50 mln from a 9+.ank consortium o E``55 in estigations .# -S and -[ regulators into unusual fluctuations in co%%er %rices o !arch E``65 Sumitomo disco ered that a statement from a foreign .ank did not match its records o >amanaka was /ailed for a #ears, Sumitomo %aid a fine of ZE50 mln in the -S and Za mln in the -[, !errill L#nch %aid a fine of ZE5 mln in the -S and ZE0 mln in the -[, o Sumitomo filed suits against !organ and other .anks in assisting the illegal trades S'ecifics of +R Com%an#+s%ecific >ard to "uantif# 4n erse relation .etween E6loss8 and 0ro.6occurrence8 o >3LS5 high fre"uenc#, low se erit# La' techni"ues o L3>S5 low fre"uenc#, high se erit# E&treme alue theor# 4ntentional 6fraud8 s unintentional 6mistakes8 4nteraction with !' and C' Cuantification of +R5 to% down s .ottom u% a%%roaches 4ndicators o [e# %erformance indicators E7g7, q wrong o%erations o [e# control indicators5 E7g7, q %re ented mistakes o [e# risk indicators 3orecast $' .ased on %erformance and control indicators 1nal#sis of 0:L olatilit# une&%lained .# !' and C' Causal models o !easure losses using conditional %ro.a.ilities Distri.ution of 0:L Mana)e%ent of +R 4nternal control s#stem o Teneral %olic# .# to% management o 1ssessment of risks o Control %rocedures 4nternal5 no conflicts of interests, dou.le checking, a%%ro ing access E&ternal5 confirmation, audit o Current monitoring 3inancial transactions s#stem o 3ront+office5 (the face of the com%an#) o =ack+office5 e&ecution of the deals, accounting G2

3inancial 'isk !anagement, NES 2006/9

o !iddle+office5 in%ut data, %re%are %a%ers, e aluate risks 4nformation s#stem5 o Data .ases, software o Securit#, aggregation, interaction

Ne$ 3asel a)ree%ent2 reser-es for +R =asic indicator a%%roach 6=4185 $'C H MT4 o T45 gross income, F# a g o M5 reser ation coefficient, E5B *he standardi;ed a%%roach 6*S185 $'C H Ii Ni T4i o a std directions5 cor%orate finance, trading o%erations, %a#ments 6all EaB8, retail 6E2B8 and commercial .anking 6E5B8, intermediation 6E5B8, asset management, .rokerage 6E2B8 o 1lternati e standardi;ed a%%roach5 loans and ad ances instead of T4 1d anced measurement a%%roaches 61!18 o 4nternal measurement a%%roach 64!185 $'C H Ii xi ELi E&%ected losses instead of T4 .ased on 0D, LTD, and correlations Can .e ad/usted for risk %rofile inde& 6'048 o Loss distri.ution a%%roach 6LD185 $'C H Ii $La'i o Scorecard a%%roach o -% to 20B of $' e&%osure can .e insured S'ecial ris!s !odel risk o @rong model o !issing risk factor o 4n%uts Low li"uidit# Legal risk o Standardi;ation E``25 4SD1 !aster agreement 1ccounting risk o !arking+to+market Low li"uidit# o >edging s s%eculati e o Swa%s o *a&ation

4ntegrated risk+management
Recent de-elo'%ents 4ncrease in olatilit# after E`9F Tlo.ali;ation Deregulation >uge growth of the 6e&otic8 deri ati es market >igher olume of the off+.alance 6deri ati es8 o%erations Securiti;ation *echnological %rogress o Electronic trading s#stems GF

3inancial 'isk !anagement, NES 2006/9

o 0rogram trading Conclusions5 o 1ggregation of risks 4m%ortance of the enter%rise+wide '! Need unified framework o >igher s#stemic risks o >igher o%erational risks

RAR+C 6=ankers *rust, end of 90s8 !ost %o%ular risk+ad/usted %erformance measure o $ther5 '$'1C, '1'$'1C '1'$C H CEarnings _ E6Loss8D / 'C o Earnings5 %rofit net of all ta&es and e&%enses o E&%ected losses5 C'5 f60D, CE, ''8 !'5 .ased on La' models o 'isk ca%ital5 reser es co ering losses with gi en %ro. for gi en hori;on Can .e ad/usted with stress testing >ori;on5 usuall# annual o *rade+off .etween !' and C' 4dentification of risks o -suall#5 !', C', and $' o 1dditional5 .usiness, e ent, .alance risks 1ggregation o Standard5 'C H !'C J C'C J $'C o !onte Carlo A''lications of RAR+C Enter%rise le el5 o E aluate efficienc# of work 6.ackward+looking8 o $%timal ca%ital distri.ution 6forward+looking8 o 4nformation for the outside world 6shareholders, regulators, rating agencies8 o !anagerial com%ensation Criti"ue o Common .ottom+u% a%%roach o =ased on total risk 6contrar# to C10!8 o 4na%%lica.le to risk+free instruments 6unless im%ose %ositi e reser es8 Re)ulation of *an!s *he standardi;ed framework s internal models o 4nternal models should satisf# certain criteria and .e a%%ro ed .# C= =asel Ca%ital 1ccord 6=4S 4, E`aa8 o Differentiate the C' e&%osures *he E``5 amendment o 4ncor%orate !' *he new =asel Ca%ital 1ccord 6=4S 44, 200F8 o 4ntegrate !', C', and $'

GG