ontinuing Education Part 1 : Risk m et hodolo g i es isk Trees in the Woods By Peter D avies Sailfish Systems With multiple risk methodologie s to choo s e from, treasurers should know what each does and then apply that most appropriate within the context of their company's business objectives. I n g oing public with i ts RiskMetrics methodology, JP Morgan h as put a very bright spot li ght on risk eva lu at i on methods. As attractive as Ri skMet ri cs is , particu l a rl y to corpora t e treasurers, there are other m et hodolo g i es that treasurers s hould be awa r e of. This overview of the three primary t ypes of risk m eas ur e m e nt determin istic, probabilistic, a nd a n a lyti ca l s hould h e lp treasury ri sk m a n agers to quickly identify th e trees a lt e rn ate r i sk measurement m et h od o lo g i es) a nd keep their eyes o n the forest ( m anag in g the r i sks aris in g from t h e ir compa n y's business ob j ec ti ves). eterministic methods The deterministic m e thod o lo g i es in c lu d e scenarios, se nsitivi ty, th e Greeks (de lt a, ga mm a, e t c . ), a nd stress test in g . Th ese methods are those most w id e l y s upport ed in traditional trading systems as they are ofte n u sed by f in anc i a l traders. Deterministic risk measurement meth- ods require those employing them to specify what they expect future mar- ket rates (scenarios) to be and value their positions accordingly. The risk measured, therefore, is that arising if the expected rates indeed occur . Traditionally, deter mini st i c methods we r e used by traders to measure the risk o f p r oducts with t h e sa m e ri sk c h a r acte ri st i cs - e.g., spot FX or US 6 treasury n otes . The s im p li city of th e risk st ru cture in su c h instruments makes it easy fo r traders to judge h ow the relatively few risk factors affect in g their positions may behave- these op inion s are inh e r e nt in t h eir t r ad in g. Th e p r oblem w ith deterministic methods is that they are highly tra n s- action o r ie nt ed. They fo c u s on the p ri ci n g a nd execut i o n of a stream of individual dea l s , acco untin g for the fact that eac h o n e m ay be misvalued before i t ca n be h edged or closed. The G r eeks gained popularity, fo r exa m ple, because they a l e r t t r ade r s to m ar g in a l se nsitivi ty to mispricing a ny price var·iab l e . Deterministic methods are n ot as we ll suited to assessi n g the risk of a portfolio of d i ve r se in st rum e nts (e.g. , der i vat i ves). Thi s is part i c ul ar l y true in a co po r ate treasury co nt ext w h e r e t r eas ur ers are n ot the direct expert trader- i .e., where ope r ati n g m a n age ment r eates t h e und er l y in g positions . Ev e n very good traders can n o t assess the r easo n ab l eness of sce n a ri o s that cover many risk factors ( m a r ket r ates as we ll as comp l ex pos it i o n s) - l et a lo n e t h e positions of ot h ers. Th u s, ad ditional risk measurement m et h od o l ogies sho uld be u sed to h e lp man age ri sk beyond th e tactical l eve l. Even w ith tactical risks, h owever, trade r s should n ot be l eft to manage their o w n positions. Stress-testing, hi g hli g ht ed in the G-30 recommenda tions, i s a n important part of a n over a ll risk m easurem e nt po li cy . I t l ooks at the extre m es, t h e unthinkables, a nd the in t ent i o n a ll y irr at i ona l. It balances th e trader ' s m e nt a li ty that positions can a l ways be traded o ut of v n in ext r eme st r ess. Management can take so m e comfort in knowing what kinds of extre m e markets w ill h ave the mo st dire co n seq u e n ces for their positions . Probabilistic methods Th e p r obab il i st i c methods includ e: JP Morgan ' s RiskMetrics, Monte Ca rl o s imul atio n , and historical s imul at i o n . Ge n er a ll y speaki n g, probabilistic methodologies forgo the ce rt a inty of determi ni stic methods a nd measure risks r esu ltin g from a n uncertain range of rates o r prices. The key to mo st probabi I s ti c methodologies i s the distribution func tion. I t describes the r a n ge of uncer ta inl y in the future rate or price vari ab l e being dealt with . O n ce a distrib u ti o n fu n ctio n is c h osen , a computer can fi ll in the ac t u a l sa m p l es w ithin the distribution w ith as much detail as de m anded. While prob a bilistic methods eliminate the need for traders to make assumptions about future rates/price s they do require statisticians to make assumptions about how future rates/ prices are distributed. The power of the distribution func tion is that a si n g l e function ca n embody many details, e.g., a n y p r oba b ility fo r any r a t e, thereby r ed u c in g t h e numb e r of c hoi ces or assumpt ion s n eeded. Un fort un a t e ly , th e r e m a inin g c h o i ces a r e much m or e obscure. Assumptions in probabilistic method o l og i es require a familiarity with sta t i st i ca l descriptions of a n in creas in g l y comp l ex a nd dynamic f i nanc i al world-and their fa llibili ty. There i s a l so a second dimension w h en probabi I s t i c methods are used to m ea s ur e the risk in a portfolio of positions. Even i f a n app ropri ate distri bution funct i o n for a ll the variables affecting the a portfolio can be speci fied , e c h di s tribution fu n c tion i s merely defining t h e r a n ge of expected future va r i ab l es affect in g eac h pos i tion. To accu r ate l y m eas ur e the r i sk in the portfolio, stat i st i c i a n s must a ls o desc ri be the r e l atio n s hip between these var i ab l es. Th e relationship b etwee n r ate/p ri ce mov e m e nt s i s usu a y in terms of cova ri a n ce. Th e covar i a n ce approach quantifies these r e l ation s hi ps by taking hi storica l data samp l es a nd ca l cu l at in g t h e distributions and re l at i o n s hi ps from past obse r vations . Covar i models are r i ve beca u se t h ey e n ab l e r i sk measurers to r ed u ce a l ot of sa mp l e data in to a few In te rn at i o n a l Tr eas ur e r /Nov e m be r 28, 1 994