Economc and Fnancal Report 2002/02 Hedge accountng wthn IAS39 Alessandro Ross, Gudo Bchsao and Francesca Campolongo Economc and Fnancal Studes European Investment Bank 00, boulevard Konrad Adenauer L-2950 Luxembourg Emal: nfoefs@eb.org Notes Alessandro Ross (alessandro.ross@rc.t) and Francesca Canpolongo (francesca.campolongo@rc.t) work for the Jont Research Centre of the European Commsson n Ispra (VA), Italy. Gudo Bchsao (bchsao@eb.org) s wth the Asset-Lablty Management dvson of the European Investment Bank. ECONOMIC AND FINANCIAL REPORTS are prelmnary materal crculated to stmulate dscusson and crtcal comment. Quotaton of materal n the Reports should be cleared wth the author or authors. The vews expressed are those of the ndvdual authors, and do not necessarly reflect the poston of the EIB. Indvdual copes of the Reports may be obtaned free of charge by wrtng to the above address or on-lne from www.eb.org/efs/pubs.htm
Abstract Ths work proposes an accountng calculaton scheme for hedgng swaps based on the requrements lsted under Internatonal Accountng Statement (IAS) 39. In partcular we developed a procedure that asssts rsk managers n the dentfcaton of the hedgng effcency between a group of loans (or bonds) and swaps held n a bank portfolo qualfyng for hedge accountng. The proposed scheme ams at assocatng to any gven swap of the bank portfolo, a certan collecton of loans (or bonds) whose rsk exposures offset each other. The fnal result s the constructon of a number of hedges that are effectve accordng to IAS 39. 2
Hedge accountng wthn IAS39. Why IAS 39 In past years addtonal attenton was devoted to the accountng standards related to fnancal nstruments. Wth the ncreased sophstcaton of captal markets and the use of dervatves, the need for accountng rules to ensure transparency n fnancal statements through adequate dsclosure of postons and exposures becomes crucal. Accountng rules need to be standardsed, so that the performance and soldty of fnancal nsttutons can be assessed and compared n a coherent way through the fnancal statements. In ths ven, regulatory bodes n dfferent countres have been workng to establsh standards for fnancal accountng and dsclosure. The need of proper accountng standards was felt even more ntensvely n February 995, when the bankng world was shaken by large-scale falures such as the Barngs Bank collapse. Accountng standard organsatons realsed that the exstng hstorcal cost conventon was unsatsfactory for fnancal nstruments gvng no ndcaton of the nherent rsks. The most mportant actors n the development of accountng standards have been the Internatonal Accountng Standard Commttee (IASC) wth ts IAS39 standard and the US Fnancal Accountng Standards Board (FASB) wth the FAS33 standard. The two standards have substantal overlap; the most sgnfcant dfference s that the FAS33 standard s lmted to hedge accountng whereas IAS39 also covers other accountng areas. IAS39 was ssued n March 999 and s effectve snce January 200 n those countres the Parlaments of whch have ntegrated the standard n the related natonal laws. The standard sets requrements for a wde range of nstruments and, for the frst tme, dscplnes the recognton, de-recognton and measurement of dervatves. In general, IAS39 requres all dervatves to be recognsed on the balance sheet and measured at far value. Ths s a man change wth respect to prevous practces where dervatves used for hedgng purposes were often off balance sheet. The requrement that all dervatves are vsble on the balance sheet ams to gve transparency and nsght nto the true exposure of a fnancal nsttuton or company. Profts and losses arsng from changes n far values of dervatves are recorded n the ncome statement. Addtonally, dervatves used to hedge a group of assets or labltes could qualfy for hedge accountng. Hedge accountng apples, however, only when strct hedge effectveness crtera are satsfed. By effectveness t s meant the degree of correlaton between changes n far value of the exposure and changes n far value of the desgnated 3
hedge. If hedge effectveness crtera are fulflled then hedge accountng can be appled wth the consequence that both the hedgng dervatves and the hedged group of assets or labltes are recorded at far value but wth opposte sgn. Measurng the hedge effectveness s therefore a crucal part of the IAS39 standard. One of the most mportant requrements for good management s to be able to dentfy whch hedgng strateges qualfy for hedge accountng under IAS39 and whch not. Accomplshment of ths task s not trval and requres the assstance of approprate procedures, especally because the lmts mposed by IAS39 on the use of hedge accountng are strct even for hedges that are economcally effectve. Ths work focuses on the concepton of a methodology that s effcent n selectng portfolos of loans (or bonds) and nterest rate swaps that fulfl hedge effectveness crtera and therefore qualfy for hedge accountng. The paper s structured as follows. Secton 2 sets the problem. Secton 3 states the obectve and explans the leadng features of the procedure, whereas Secton 4 descrbes ts functonng by means of an example. Secton 5 concludes. 2. IAS39 requrements As the use of dervatves has become more and more ntensve, concern about ther accountng has ncreased. The man concern s that dervatves are tradtonally off balance sheet tems and there s lttle dsclosure about the rsks mpled by ther use. IAS39 s an effort along the drecton of an mproved transparency by means of strct rules for the accountng of dervatves. The problem can be llustrated by the followng example. A fxed rate debt s converted nto a varable nterest rate va an nterest rate swap. The debt and the swap are consdered together as a synthetc varable rate borrowng, and the amount recevable or payable under the swap s used to adust the nterest rate payable on the debt. Before the ntroducton of IAS39 the accountng of these nstruments was straghtforward: the swap, provded t was perfectly matchng the debt oblgaton, was dsclosed as part of the terms of the debt and was recorded at ts nomnal cost together wth the debt. However, ths way of accountng makes t dffcult to determne the underlyng economcs and to assess and manage the rsk ncurred through dervatves. Hence the need for dsclosure and the logcal choce to separate dervatves from the underlyng host contract recordng each tem at ts far value to make the rsk more vsble. An excepton to the general rule can be made when hedge accountng s applcable but only n a specal case. In fact, hedge accountng s applcable n two cases:. the dervatve s mrrorng the cash flows nherent to the asset or lablty t s hedgng 4
2. the dervatve s closely mrrorng the exposure profle of a group of assets or labltes In the frst case, the so-called short cut method s used;.e. the hedged tem far value s consdered to be equal to the far value of the hedgng dervatve and no effect s generated n the P&L by the movement n the market condtons. In the second case, the effectveness of the hedge s evaluated and whenever t s not perfect, the porton of neffectveness (whch s lmted by strct rules) s recorded n the P&L. When hedge accountng does not apply, the dervatve s recorded at ts far value, whereas assets (typcally loans) and labltes (typcally bonds) are recorded at ther hstorcal cost leadng to a clear assessment of the dervatve exposure. In summary, the general prncple set by IAS 39, regardless the applcablty of hedge accountng or not, s to ensure dsclosure of all type of rsks assocated wth a dervatve, ether they be related to the dervatve tself or to the neffectveness of a hedge. Ths can be avoded only when the hedge can be regarded as 'perfect' and the short cut method s applcable. IAS39 establshes requrements governng whenever transactons entered nto for hedgng purposes qualfy for hedge accountng. The new requrements are restrctve reducng the transactons for whch hedge accountng s appled. Hedge crtera nclude: () Both the hedged tem and the hedgng nstrument should be clearly dentfed and documented. Management must document exactly what s the hedged rsk and how t wll assess the effectveness of the hedge. (2) The hedge must be effectve: at the ncepton of the hedge, the mpact of the hedged rsk on the hedged and on the hedgng tem must ''almost fully'' offset; subsequently, effectveness must be tested regularly throughout ts lfe. Requrement (2) clams that, to be classfed as effectve, the hedge does not have to be perfect. The constrant s rather that the hedge s expected to be: () Hghly effectve at ncepton,.e. changes n far values should almost fully offset, and () Effectve n practce throughout the lfe of the hedgng relatonshp.e. the rato of the change n far value of the hedged tem and the hedgng tem must reman wthn a range of 80% to 25%. If durng ts lfe the hedgng relatonshp fals to reman wthn the pre-set range, the dervatve wll be accounted for at ts mark-to-market whereas the assets or labltes at hstorcal cost. A thrd hedge accountng constrant apples to the hedge of an expected future transacton. In ths case the transacton beng hedged must be 'hghly probable', whch means ts tmng can be forecast relably wthn three months perod (see []). However, as n ths work we have not consdered the problem of hedgng expected future transactons, n the rest of ths document we wll focus only on the frst two requrements. 5
Ths work s based on a common example: nterest rate swaps hedgng a group of loans. In partcular, we developed and mplemented a procedure that asssts practtoners n the montorng of the hedgng relatonshps between swaps and loans qualfyng for hedge accountng. IAS39 recognses three knd of hedgng relatonshps, accordng to the type of rsk beng hedged and to the source of the exposure: far value hedge, cash flow hedge, and hedges of a net nvestment n a foregn entty. A detaled descrpton of these hedgng relatonshps s not n the scope of ths analyss (whch s restrcted to the far value hedge) and can be found n [2] (page 43). In the far value hedge the rsk beng hedged s a change n the far value of a recognsed asset or lablty (due n our case to a movement n nterest rates) that wll affect the ncome statement. The followng secton llustrates a procedure whose goal s to dentfy as many as possble hedgng relatonshps meetng the above mentoned accountng constrants necessary to qualfy for hedge accountng and to maxmse the effectveness of the hedge. Accordng to the requrement of IAS39, a hedgng relatonshp s defned for the entre lfe of the hedgng nstruments. The procedure assocates to any gven swap a combnaton of loans (or bonds) so that the resultng hedged portfolo s effectve (pont (2) above). In practce, the procedure ams at fulfllng the effectveness crtera at ncepton (frst constrant) leavng the montorng of the effectveness of the hedge durng ts lfe to an ex-post calculaton tme. 3. The procedure In the followng secton, we wll use the IAS39 termnology, and we wll refer to a portfolo of a hedged and a hedgng tem as to a ''hedge''. As outlned above, the goal of ths work s to buld hedges that qualfy for hedge accountng. The dea s to search, for any gven swap, the combnaton of loans (bonds) whose senstvty to changes n the yeld curve s of the ''same type'', but opposte n sgn. By the ''same type'' we mean that n any gven pont of the yeld curve the senstvtes of the two nstruments are of the same magntude. If ths were the case the hedge would be nsenstve to changes n the yeld curve. Therefore, we would expect ts net present value to be almost unaffected by these changes. The selecton of the best combnaton of exstng loans (bonds) gven a swap s based on the prncple that at the tme IAS39 s appled to a balance sheet, a bank should dentfy n ts exstng balance sheet, the exstng transactons (and related dervatves) for whch hedge accountng s sought. Typcally, banks were extensvely usng so-called ALM swaps to hedge the entre balance sheet exposure,.e. plan vanlla swaps whch were reducng the total senstvty of the balance sheet to nterest rate movements. In ths nstance, the bank s confronted wth the problem that a unque relatonshp between hedgng dervatves and hedged assets or labltes s not avalable. Therefore, the bank should extract from ts pool of loans or bonds the combnaton whch s best hedged from the plan vanlla swaps entered nto for ALM reasons. 6
Let S denote the swap and L, L2,..., Lk, a set of loans. The hedged tem can be wrtten as x L where x ndcates the porton of loan L used. Denote by MMkt(S) and MMkt( x L ) the actual swap and hedged tem mark to market, and by MMkt (S) and MMkt x L ) the mark to market values as computed n the th scenaro (=,2,3) ( where = corresponds to a hypothetcal shft of the yeld curve of 00 bps up, =2 to an hypothetcal shft of 00 bps down, and the =3 to an nverson of the yeld curve. Under IAS 39, the hedge wll qualfy for hedge accountng f condtons () and () stated n the prevous secton are fulflled. Condton () (.e. the rsks of the hedged and hedgng tems at ncepton almost fully offset), requres that the varaton of the mark to market of the swap and of the hedgng tem, as a consequence of changes n the yeld curve, are roughly of the same magntude. Ths translates nto the followng relatonshp holdng at ntal tme t 0, and for each hypothetcal scenaro =,2,3. k = MMkt ( S) MMkt( S) lb < < ub MMkt ( x L ) MMkt( k = x L ) =,2,3, () where lb and ub are pre-establshed lower and upper bounds. A hedge fulfllng relatonshp () s sad to be effectve at ncepton. As tme passes, the mark to market values of the consdered tems change and the above relatonshp can be affected by: changes n the yeld curve; occurrence of maturng cash flows ether n the swap or n the group of loans (bonds). A change n the relatonshp s accepted by IAS39 whch requres that the hedge must reman ''effectve n practce'' durng all ts lfe. The meanng of ''effectve n practce'' translates nto the followng relatonshp holdng at any tme t > t 0 : The same methodology can be appled to bonds. 7
k = MMkt ( S) MMkt( S) + C lb < < ub MMkt ( x L ) MMkt( k = S H x L ) + C (2) where cashflows are taken nto account by addng the quanttes C and C that represent the total amount of cash flows matured between t 0 and t both on the swap and the hedged tem sde. lb and ub are set respectvely to 80% and 25%. When relatonshp (2) ceases to hold, the hedge becomes neffectve and t has to be dscontnued. The procedure s successful n buldng hedges that are ntally effectve. However, the montorng of the effectveness of the relatonshp (2) durng the lfe of the hedge s left to the rsk manager and s not handled by the procedure. For each loan and swap the vector of senstvtes η of dmenson m+ (where m s the number of ponts n tme for whch observatons on the yeld curve are avalable) s calculated. For =,2,...,m, the th component of η, say η, represents the senstvty of the mark to market value to changes n the th pont of the yeld curve. The m components are known as key rate duratons (see [3]). The (m+)th component represents the total senstvty,.e. the sum of the frst m components, multpled by a factor P: m η = η, η2,, ηm, Pη,, = where P s chosen by the user. The role of P wll be clarfed later. In order to reduce the computatonal burden, for any swap S, a restrcted number of K loans that may be ''more sutable'' as hedged tems are selected. The ''more sutable'' loans are those whose vector of senstvtes η s smlar (.e. exhbtng senstvty n the same ponts of the yeld curve) and of opposte sgn compared to the swap. K s chosen by the user. Wthn the K-subset, the ''best '' combnaton of k loans wll then be chosen. The procedure begns by dentfyng the K-subset. For ths we defne the Tme Average Senstvty (TAS) ndex, whch s gven by TAS( Y) m = = m t η ( Y ) η ( Y ) = where Y={S,L}, and η ( Y ) s the senstvty of Y to the th pont t of the yeld curve. The K loans chosen are those exhbtng the ''closest '' dstance from the swap n terms of TAS where the dstance s defned as TAS( S) TAS( L). The second step ams at fndng the k-combnaton of loans, wthn the K-subset, that ''best'' hedge the swap S. In other words, the goal s to fnd the hedge whose loans are S H 8
wthn the K-subset and whose effectveness at ncepton s maxmsed. In our framework ths corresponds to mnmse the dstance between the two vectors of senstvtes. Ths s acheved by two subsequent steps. Frst, gven k loans, the procedure searches for the sequence x, x2,..., xk that mnmses the obectve functon k = 2 η( S ) x η( L) (3) under the constrans () () 0 x d, a ub k C x a 3 = lb3 where d s the proporton of the th loan stll avalable (not prevously used for hedgng purposes), a = MMkt3 ( S) MMkt( s) and C = MMkt3 ( L ) MMkt( L ). The subscrpt 3 denotes the curve nverson scenaro. 2 The mnmsaton of the target functon s appled to all the k over K combnatons of loans and the mnma are stored together wth the correspondng coeffcents. Then, as a second step, the lnear combnaton of loans for whch the obectve functon shows a global mnmum s selected. Loans and swap form the hedge that s a canddate for hedge accountng. A dscusson about the choce of the target functon (3) and of the constrans s now requred. Whle the choce of the metrc and that of the frst m entres of the vector of senstvtes s qute ntutve, the last entry, the total senstvty tmes the parameter P deserves some comments. In general there s a trade off between hedgng the swap aganst parallel shfts and curve nverson. Hedges, whch are robust to parallel shfts of the yeld curve, are senstve to curve nverson and vce-versa. Snce the mnmsaton of total senstvty amounts to mmunse the hedge aganst parallel shfts, hgher values of P ental a better hedge aganst that scenaro, but lkely worst results n terms of curve nverson. The opposte holds for lower values of P. Then the user may act on P to establsh the desred portfolo robustness n the frst two or n the thrd scenaro. Constran () s a natural requrement, whereas constran () has been ntroduced to force () to be fulflled n the curve nverson scenaro. Furthermore t speeds up the procedure. 2 The mnmsaton problem defned here s equvalent to the Kuhn-Tucker problem that searches for a 2 mnmum of a quadratc functon f ( x ) = Ax b under certan constrans for x. The soluton s found by usng the Fortran routne E04NCF provded wthn the Nag Lbrary (mark 9). 9
It s worth notng that the obectve functon has been defned ndependently from the scenaros used to measure hedge effectveness. Hence we may expect the procedure to perform well even under dfferent scenaros. The algorthm s terated for all the avalable swaps. Swaps are processed gong from the hghest to the lowest mark to market value. A dfferent orderng may be lkely to produce dfferent results. It s mportant to stress that a hedge selected by the procedure s a canddate for hedge accountng but not necessarly elgble. In prncple the best hedge avalable may stll be neffectve n one or more scenaros and therefore unsutable for hedge accountng. The followng secton shows an example of a hedge that s ntally effectve and stays ''effectve n practce'' throughout ts lfe. 4. Example Here we consder a bullet payer swap, whose man features are reported n the table below. 3 Pr Rate TTM TAS Tot. Sens. MMkk 28 5.82 4.33 4.3 82,593-4.7 Tme to maturty (TTM) and TAS are expressed n years, the prncpal (Pr) and the mark to market (MMkt) n mllon Euros. The swap s hedged by usng at most k=7 loans. The number of loans avalable s 428 (amortzng and bullet). The actual yeld curve s shown n Fgure. 3 The data used n the example were provded by the European Investment Bank. 0
The senstvtes of the swap n each pont of the yeld curve are plotted n Fgure 2.
As evdent from the pcture the swap exhbts greater senstvty ust before ts expry date. The parameters of the routne are set to K=20 and P=0. Portfolo total senstvty and hedge effectveness at ncepton for each of the three scenaros (n percentage ponts) are reported below: Tot. Sens. +00bps -00bps Curve Inv. -28.37 99.93 00.00 0.93 The hedge has proved to be almost perfectly effectve n scenaro and 2 and very good n scenaro 3 and therefore qualfes for hedge accountng. The hedgng tem s composed by loans wth the followng features: Id % Type * Pr TTM TAS Rate Fr 00 B 60. 4.42 4.33 5. y 2 00 B 2.5 4.4 4.26 4.33 y 3 74.2 A 76.2 4.75 4.27 - - 4 00 B 5.5 4.42 4.32 5.42 y 5 00 B 8.76 4.42 4.34 4.3 y 6 00 B.24 4.42 4.34 4.3 y 7 00 B 33.9 4.42 4.34 4.33 y One may observe that bullet loans are entrely taken. Ths s because all these loans have tme to maturty almost equal to that of the swap but smaller prncpal. Therefore, ther senstvtes are almost concdent n tme, opposte n sgn and smaller n magntude. The amortzng loan, whose senstvty s spread over tme, s fnally selected snce the amount of bullet loans was not suffcent to offset the swap senstvty. The senstvty of the resultng hedge s shown n Fgure 3. * A = ammortsng, B = bullet 2
Note that the hedged tem has consderably reduced (by a factor of ten!) the hgh senstvty peak dsplayed by the swap. As tme passes by the hedge tends to lose ts effectveness. IAS 39 takes ths nto account, but t requres the hedge to reman "effectve n practce" through tme. In ths example we have chosen four dates to verfy hedge effectveness n practce by computng the ratos gven n (2). In each date we arbtrarly mpose a yeld curve evoluton. The yeld curves are shown n Fgure 4. 3
The effectveness ratos are shown below Tme +00bps -00bps Curve nv 2 m 99.35 99.49 99.38 6 m 93.49 93.93 93.60 28 m 96.02 96.25 95.86 40 m 92.7 92.94 93.8 Although lower than at ncepton, the ratos fall wthn the pre-establshed bounds and the hedge remans effectve. The choce of the ponts n tme at whch ratos are computed s arbtrary. However our choce focussed on those date that are expected to be crtcal, such as those mmedately after a cash flow payment from the swap sde. In these crcumstances there s a lack of balance between the cash flows stream of the swap and that of the hedgng tem that are lkely to produce a bad performance of the hedge. The fact that n these dates the ratos are wthn the pre-establshed bounds strengths our confdence n the qualty of the hedgng performance. 4
5. Concludng remarks In ths work we have proposed an algorthm dentfyng hedges nvolvng swaps and loans (or bonds), that qualfy for hedge accountng under IAS 39. The effectveness of the hedges s then montored through tme although the choce of the ponts n tme when the hedge s to be montored s left to the user. The core of the procedure s the target functon (3), whch seems to be properly desgned as confrmed by the example and several other trals. Nevertheless other forms of ths functon maybe thought of. For nstance one may also take nto account the degree of correlaton among yeld rates as n [4]. A fnal remark regards the choce of the scenaros. In our vew parallel shfts and curve nverson scenaros are probably very smple cases. More realstc scenaros could be consdered, for nstance by makng use of prncpal components analyss. Identfyng two or three man factors (or components) whch drve the yeld curve evoluton one could then compute the ratos wth respect to movements of those factors. Fnally, t must be noted that hedge effectveness s nfluenced also by the varable leg of the hedgng swap. In fact even n case of a fxed leg of a swap mrrorng the cash flows of a set of loans, the mark to market of the hedgng tem and the hedged tem wll not behave at the same way. Ths effect becomes unmanageable f the second leg of the swap s not a varable rate but a fxed rate. 6. References [] Prce Water House and Coopers, Fnancal Instruments. IAS39: The Essental Gude to the Q&As. [2] Prce Water House and Coopers, Internatonal Accountng Standard, Fnancal Instruments: Understandng IAS 39. [3] Ho T.S.Y., (992): "Key Rates Duratons: Measures of Interest Rate Rsks ", The Journal of Fxed Income, pp 29-44. [4] Falkensten E. and Hanweck J. (996): "Mnmzng Bass Rsk from Nonparallel Shfts n the Yeld Curve ", The Journal of Fxed Income, Vol 6 No. 5