Deconstructing FISIM: Should Financial Risk Affect GDP? Dennis Fixler and Kim Zieschang* July 30, 2010

Transcription

1 1 Deconsrucng FISIM: Should Fnancal Rsk Affec GDP? Denns Fxler and Km Zeschang* July 30, 2010 *U.S. Bureau of Economc Analyss and Inernaonal Moneary Fund, respecvely. The vews expressed heren are hose of he auhors and should no be arbued o he Bureau of Economc Analyss or he IMF, he IMF Execuve Board, or IMF managemen.

2 2 I. INTRODUCTION The Sysem of Naonal Accouns (SNA) s he nernaonal sascal sandard ha underpns closely wached macroeconomc ndcaors such as gross domesc produc (GDP). Fnancal servces measuremen has become opcal n recen years, and he SNA s mehodology for calculang he oupu of he fnancal servces ndusry s recevng closer scruny, no leas because of he recen fnancal crss. The reamen of fnancal rsk whn hs mehodology has become an mporan ssue under dscusson n he naonal accounng communy and among users of hese daa. The SNA s approach depends crcally on how a reference rae of neres s deermned, as hs affecs he esmae of he curren value of oal oupu of he servces of banks and oher fnancal nermedaon nsuons. I also affecs he dsrbuon of hose servces beween deposors and borrowers, and her dsrbuon across economc secors. An ssue has arsen wheher he reference rae for loans should nclude rsk prema and, by mplcaon, wheher he measure of he curren value of cred servces oupu should exclude all or par of hese prema. More han ffeen years ago, he 1993 verson of he SNA nroduced he concep of fnancal nermedaon servces ndrecly measured (FISIM) as he preferred approach o measurng he value of ndrecly charged fnancal servces producon. Pu smply, FISIM for an asse class such as loans s he neres earned on he asse class less he ncome foregone a a reference rae of neres for he asse class. FISIM for a lably class such as deposs s he ncome ha would be earned a a reference rae of neres on he funds acqured less he neres cos of hose funds. FISIM s broadly conssen wh he Barne (1978)-Donovan (1978) user cos of money heory of fnancal servces producon and consumpon, as foreshadowed by Dewer (1974). Because he 1993 SNA allows for nonzero value added and fnal consumpon of ndrecly measured fnancal servces, FISIM affecs GDP for counres complng naonal accouns under he 1993 SNA. 1 The predecessor sandard, he 1968 SNA, allowed for ndrecly measured fnancal servces producon bu, by convenon, allocaed all of hs producon o nermedae consumpon; hus ndrecly measured fnancal servces made no conrbuon o GDP. The newes, 2008 verson of he SNA no only recommends bu requres he use of FISIM for loans and deposs, allocaed o boh nermedae and fnal consumpon, and hus affecng GDP. An mporan ssue n mplemenng FISIM on whch full consensus has no been reached s how he reference rae should be deermned. The 1993 SNA suggesed he reference rae be a very shor duraon, essenally rskless rae, lke he nerbank rae. The 2008 SNA 1 Counres are allowed by he 1993 SNA o op ou of allocang par of FISIM o fnal consumpon, however, so no all 1993 SNA counres may, n fac, follow he frs-bes mehodology (1993 SNA, paragraph ).

3 3 (paragraph 6.166) connues o say he nerbank rae may be suable, bu also says The reference rae should reflec he rsk and maury srucure of deposs and loans. The 2008 SNA hus denfes wo subsdary consderaons n deermnng he reference rae, he degree o whch should reflec he level of rsk of he nsuon s porfolo of asses and lables, and he degree o whch should reflec he resdual maures of he asse as well as lably classes o whch s appled. The SNA s language (ncludng oher ex n he same chaper), mples ha here s a sngle reference rae appled o all ypes of deposs and loans, bu ha hs rae should reflec he overall rsk and maury of deposs and loans n some average sense. As a counerpon, Wang, Basu, and Fernald (2004) and Basu, Inklaar, and Wang (2008), among ohers, argue for a consellaon of reference raes fully reflecng he rsk and maury prema of each asse class. Snce he reference rae s subraced from he loan rae n calculang ndrecly measured cred servces, hs effecvely elmnaes rsk and maury remuneraon from FISIM. In hs paper, we observe ha banks ndrecly measured fnancal servces comprse no only paymen processng and loan orgnaon acves, bu also acves mgang nsolvency rsk o deposors, loan defaul rsk, and he rsk of losses arsng from neres ncome falng o cover neres expense because of dfferen maury profles beween asses and lables. Rsk ha nsuons canno mgae, eher hrough borrower screenng, loan monorng, and provsonng, or hrough buyng nsurance n he form of depos nsurance as well as cred defaul and neres rae swaps, s borne by he owners and unsecured credors. In effec, he rsk bearng of owners and credors, wheher he rsk arses from he nsuon s own fnancal produc porfolo, or from agreemens o bear rsk ransferred from oher nsuons hrough fnancal dervaves, s a prmary servce ha mus be compensaed o secure capal for he enerprse. Owners and credors rsk prema are bul no operang surplus. Owners requre a reurn ha ncludes a premum for rskng her capal n he enerprse and credors receve rsk prema n he neres raes on unsecured deposs of, bonds ssued by, and loans owed by fnancal enerprses. Whou coverng hese prema hrough proceeds from he sale of fnancal servces, ncludng ndrecly measured fnancal servces, fnancal nsuons could no secure he fnancal capal hey need o operae. Examnng US daa, we fnd ha he sum of esmaed depos, erm, and cred defaul rsk nsurance servces are he preponderance of FISIM, srongly suggesng ha he greaer par of loan rsk prema are coverng eher rsk mgaon or rsk bearng coss. As we have argued above, rsk mgaon s nexrcably lnked o he rsk managemen acves o whch he frm devoes nermedae npus as well as labor, capal and fnancal resources. Accordngly he rsk prema are no merely ransfers from borrowers o fnancal nsuons. We hnk hs rases some dffcul quesons for he advocaes of removng rsk remuneraon from FISIM.

4 4 Alhough we fnd lle evdence for purgng rsk prema from FISIM, we do fnd ha he SNA s valuaon of loans a conrac value on he balance shee s napproprae for he calculaon of fnancal servces oupu. 2 Recordng loans a conrac (book) value s one of he excepons he 2008 SNA makes o he marke valuaon prncple for fnancal asses. 3 We propose nsead ha FISIM n curren prces be calculaed ncorporang he marke or far value of loans even hough loans on he SNA balance shees are a book value. Ths approach would be compable wh exsng 2008 SNA gudance on calculaon of FISIM, whou requrng a subsanve revson of he curren naonal accounng sandard. Fnancal nsuons manage rsk nernally hrough provsonng (self nsurance), or va fnancal dervave conracs (rsk ransfer), prncpally neres rae and cred defaul swaps. The growh n he use of fnancal dervaves reflecs he subsanal ousourcng of erm rsk managemen snce a he 1980s 4 and defaul rsk managemen snce he 1990s. 5 The 2008 SNA recognzes he fees charged o se up dervave conracs as paymen for servces, whch are recorded as oupu from he producers and as nermedae consumpon o he users of hese servces. However, possble servce charge componens bul no he ne selemen paymens beween pares over he duraon of hese conracs are recognzed n he 2008 SNA (paragraph ) bu are reaed along wh ransacons n fnancal dervaves because of he dffculy n dsngushng he servce elemen. Our purpose here s no o challenge he 2008 reamen of fnancal servces assocaed wh fnancal dervaves, bu o use nformaon on he rsk prema n dervave conracs o esmae he marke value of he rsk managemen servces fnancal nsuons are producng for hemselves. To produce hese esmaes, we rea hese rsk managemen servces as non-lfe nsurance produced for own use. 2 We do no challenge n hs paper he SNA s recordng of loans a conrac value on he balance shee, even hough we argue ha FISIM should be based on marke value. The wo can coexs whn he same sysem. On he 2008 SNA balance shee, revaluaons of loans appear n he form of dscouns or prema o conrac value when realzed hrough, e.g., change of ownershp. 3 The excepon s made regardng balance shee posons n loans because loans are consdered no o be markeable (even f ransacons n loans are recognzed), and because here s neres n rackng he conracual lably of he owng un, and o ensure symmerc recordng among economc agens. The fnancal accoun deals wh ransacons n loans a marke value by recordng a realzed holdng gan/loss for he seller and a realzed holdng loss/gan o he buyer. The loan s ransferred a conrac value from he balance shee of he sellng o he buyng un. There are no unrealzed gans/losses recorded on loans. The SNA also records posons n deposs and n oher accouns recevable/payable n nomnal erms. 4 Ludwg (1993), Chaper 2, Hsory of he Swap Marke. 5 US banks were usng hese nsrumens as early as See US Offce of Comproller of he Currency (1996).

5 5 We also wan o emphasze ha our focus here s on he curren prce oupu of ndrecly measured fnancal servces oupu n he naonal accouns, no he prce and volume decomposon of ha oupu. Recenly Haldane, Brennan, and Madouros (2010), among ohers, have quesoned he conrbuon of fnancal servces o GDP growh. Ths s ceranly a legmae queson. In answerng, here are wo componens of he conrbuon of he fnancal servces ndusry o GDP growh o assess: he ndusry s share n GDP and growh n volume of oupu. Whle we can hnk of refnemens, we hnk he share of fnancal servces n GDP s broadly properly measured by exsng gudelnes n he SNA, ncludng for FISIM, and ha has ndeed grown n recen years. We hnk he real ssue s wheher here has been anywhere near as sgnfcan an ncrease n he volume of oupu of fnancal servces, and, drecly relaed, wheher here has been subsanal prce nflaon n ndrecly measured fnancal servces n parcular. Regardng he role of rsk n hs nqury, here s an mporan dsncon o be made beween he mpac of rsng rsk on he ransacon value of oupu and he prce and volume decomposon of ha oupu. Rsk bearng and rsk mgaon servces mus be pad for, and rsng rsk remuneraon clearly mus ncrease he value of fnancal servces oupu. For a consumer of fnancal servces, any unnsured rsk n a ransacon reduces he qualy of he servces receved and, by mplcaon, rases he prce. Thus, ncreases n he rsk borne by consumers of fnancal servces subrac from he growh n he oupu of hose servces. Fnally, our man focus n hs paper s on rsk prema and rsk managemen servces. Alhough rsk s a facor when consderng deposs on he lably sde of he regser, for example, hrough depos nsurance prema and n managng he neres/erm rsk of varable rae deposs, s prncpal mpac on bank operaons s n loans on he asses sde. We dscuss nsurable rsk for deposs, bu our laer emprcal analyss focuses on loans. We plan o exend he emprcal analyss o nclude a fuller reamen of deposs a a laer sage. II. ACCOUNTING FOR INDIRECTLY MEASURED FINANCIAL SERVICES IN THE USER COST FRAMEWORK A. Bascs of he framework The user cos framework values he bundle of fnancal servces ha aach o he fnancal producs provded by banks. As noed n he nroducon, Donovan (1978) and Barne (1978) appled he user cos of capal concep o valung he servces provded by fnancal asses, as foreshadowed by Dewer (1974). In hs approach, he user cos of money s defned for a fnancal asse as π = h ρ

6 6 and for a fnancal lably as π = ρ h where π = he rae of user cos servce value per nomnal un of asse or lably h = he rae of holdng revenue per nomnal un asse or he rae of holdng cos per nomnal un of lably ρ = he opporuny cos rae of money, wha he SNA calls he reference rae. For loans, h s he loan rae recevable by a fnancal nsuon, and for deposs, h s he depos rae payable by a fnancal nsuon. Hancock (1985) appled he Barne-Donovan user cos of money approach o modelng he producon of fnancal servces from banks, effecvely layng a mcroeconomc foundaon for he 1993 SNA reamen of fnancal servces producon. In he same ven as Hancock s fnancal frm model (see also Fxler and Zeschang [1991]), he 2008 SNA values he ndrecly measured cred and deposor servces oupu of fnancal nsuons as Y = = π L h L + j j π D j j = h j D j ρ ( h ρ ) L + ( ρ h j ) L j j D j D j where, a me, L s he value of loans of class and D j s he value of deposs of class j. Thus, cred and deposor servce oupu n he naonal accouns s neres on loans less neres on deposs, less he cos of fnancal capal ed up n loans less deposs. 6 In he user cos formulaon above he prces of he fnancal servces conan a rsk premum, as evdenced by he dfference beween he relevan neres rae and a benchmark rae ha s aken o be a rsk-free rae. The exsence of hs premum ndcaes ha rsk bearng s nexrcably lnked o he producon of fnancal servces. For example, cred servces whose provson s necessarly ncludes a defaul rsk are ncurred along wh he rsk 6 See he appendx for he algebra ha equaes he user cos wh ne neres plus dvdends plus ren f he scope of he calculaon s he enre balance shee.

7 7 bearng he bank ncurs. I s for hs reason ha banks consume resources o manage rsk. A few excerps from he Bank of Amerca 10-K flng o he Secures and Exchange Commsson n December 2008 llusraes he pon. Our managemen governance srucure enables us o manage all major aspecs of our busness hrough our plannng and revew process ha ncludes sraegc, fnancal, assocae, cusomer and rsk plannng. We derve much of our revenue from managng rsk from cusomer ransacons for prof. In addon o qualave facors, we ulze quanave measures o opmze rsk and reward rade offs n order o acheve growh arges and fnancal objecves whle reducng he varably of earnngs and mnmzng unexpeced losses. Marke rsk s he rsk ha values of asses and lables or revenues wll be adversely affeced by changes n marke condons such as marke movemens. Ths rsk s nheren n he fnancal nsrumens assocaed wh our operaons and/or acves ncludng loans, deposs, secures, shor-erm borrowngs, long erm deb, radng accoun asses and lables, and dervaves. Marke-sensve asses and lables are generaed hrough loans and deposs assocaed wh our radonal bankng busness, cusomer and propreary radng operaons, ALM process, cred rsk mgaon acves and morgage bankng acves. We have esablshed and connually enhance conrol processes and use varous mehods o algn rsk-akng and rsk managemen hroughou our organzaon. These conrol processes and mehods are desgned around hree lnes of defense : lnes of busness, enerprse funcons and Corporae Aud. The lnes of busness are he frs lne of defense and are responsble for denfyng, quanfyng, mgang and monorng all rsks whn her lnes of busness, whle ceran enerprse-wde rsks are managed cenrally. For example, excep for radng-relaed busness acves, neres rae rsk assocaed wh our busness acves s managed cenrally as par of our ALM acves. For example, for Rsk Managemen, a senor rsk execuve s assgned o each of he lnes of busness and s responsble for he oversgh of all he rsks assocaed wh ha lne of busness. Enerprse-level rsk execuves have responsbly o develop and mplemen polces and pracces o assess and manage enerprsewde cred, marke and operaonal rsks. The above user coss were formulaed as hough all of he neres raes nvolved n he calculaon are known. Fuure neres rae values are unceran and hus confron depos akng and loan makng nsuons wh he rsk ha plans made on he bass of raes oday may no be conssen wh acual raes omorrow. Barne and Wu (2005) derve an exac expresson for he user cos of money perceved by a rsk averse household when confroned by moneary asses wh sochasc neres reurns. Under a srucural assumpon conssen

8 8 wh he capal asse prcng model (CAPM), hey show ha he user cos of a rsky asse follows he followng formula: ( Π Π ) π * = π + β * where π * = he rsk adjused user cos of asse π = he cerany user cos of asse Π * = he rsk adjused user cos of he wealh porfolo Π = he cerany user cos of he wealh porfolo β s he coeffcen of correlaon beween he gross reurn on asse and he reurn on he overall wealh porfolo (ha s, asse s bea ). The bea of a gven rsky asse s he coeffcen of correlaon of he asse s gross reurn wh he gross reurn of he enre asse porfolo. When he rsky asse s gross reurn s posvely correlaed wh he gross reurn on he full porfolo, gven ha he user cos of he full porfolo s posve, he rsk adjused asse user cos of he rsky nermedaon asse exceeds s cerany value. Rsky asses wh negave porfolo correlaons wll have rskadjused user cos raes below her cerany user cos raes. Basu, Fernald, and Wang (2004) focused on he SNA s cred servces or loan FISIM o examne wheher offcal sascs correcly accoun for rsk n measurng he cred (loan) servces oupu of banks. They argue ha he sysemac rsk prema banks receve on loan asses should be removed from her reurn n calculang ndrecly measured bank oupu. In he same ven, Basu, Inklaar, and Wang (2008) esablsh reference raes for each ype of loan by aempng o mach loan asses wh secury asses by maury and sysemac rsk. 7 They ake he dfference beween loan raes and hese mached reference raes as he value of fnancal servces, effecvely elmnang rsk prema from cred FISIM and suggesng a much lower share of bankng n GDP han currenly measured n offcal sascs. 7 See Wang, Basu, Fernald, and Wang (2004), and Basu, Inklaar, and Wang (2008). Ther specfc mehodology was o mach commercal and ndusral loans wh commercal paper asses by maury, under he assumpon ha he sysemac rsk profle of he commercal paper of a gven company maches s rsk profle as a borrower from a fnancal nsuon. The raes on commercal paper are assumed no o cover any nermedaon servces, whle he raes on loans mus cover banks nermedaon coss. Under he former assumpon parcularly, he dfference beween loan raes and commercal paper raes s aken o be he value of nermedaon servces. By defnon, excludes remuneraon for rsk and maury. A lmaon of hs mehodology was ha commercal paper s, by defnon, lmed o 270 days duraon, bu, n prncple, he mehodology could be exended o longer maures by usng corporae bond daa.

9 9 We argue ha a a concepual level hs approach denes he coss loan makng nsuons ncur n acqurng prvae nformaon abou poenal borrowers. However, even f we gran hese auhors concepual approach, her mplemenaon of exaggeraes s mpled downward mpac on FISIM. Ther emprcal approach reles on nformaon symmery beween banks (e.g., n makng commercal loans) and oher cred provders (e.g., buyers of he commercal paper of busness borrowers) ha s unlkely o hold. Banks have an advanage n knowledge of he characerscs of ndvdual borrowers and her cred rsk compared wh fnancal marke cred provders. The laer hus wll charge a hgher premum o cover her greaer perceved rsk han wll banks. Hence, a bank reference rae based on cred marke rsk prema wll be oo hgh and produce a FISIM ha s oo low, even f we were o agree ha rsk prema be excsed from FISIM (whch we do no). The mplcaon of he Basu, e al approach s apparen n he followng able of US daa whch consders he mpac of removng he mpuaons of cred and depos servces from Gross Operang Surplus whch conans he above expresson for oupu Y. Table 1. Impac of Cred and Borrower Servces on Gross Operang Surplus Gross operang surplus (bllons) Impued Deposor Servces Impued Borrower Servces Gross operang surplus, less mpued servces Gross operang surplus s from he fnance and nsurance ndusry enry of he Componens of Value Added by Indusry able. (revsed 5/25/10) hp:// Impued deposor and borrower servces are from NIPA able 7.12, Impuaons n he Naonal Income and Produc Accouns (revsed 9/3/09). hp:// Year=2008

10 10 Though he subracon of he enre ndrecly measured servce charge s exreme, s clear ha reducons n he cred and depos servces wll sgnfcanly reduce gross operang surplus and value added. In he followng secons, o derve an mplc value of rsk managemen servces ha mus be covered n value added and operang surplus, we characerze hese servces as non-lfe nsurance; ha s, fnancal corporaons can underake forms of self-nsurance usng purchased npus or, a a cos, engage n ransacons ha cede he rsk hey have ncurred. In effec, we deconsruc FISIM no erm rsk, defaul rsk, and oher accoun servces componens as well as an esmae for he ne non-lfe nsurance componen of FISIM. We fnd ha he esmaed value of nsurance servce producon agans he specfed rsks explans almos all of he marke nsurance prema. By mplcaon and conssen wh he above, elmnang rsk prema enrely would hrow ou he mplc value of he rsk managemen acves n fnancal nsuons. Our analyss hus suggess cauon n accepng he large downward adjusmens o offcal esmaes of value added from ndrecly measured fnancal servces ha are nheren n he approach of Basu, Inklaar, and Wang (2008). B. A narrow bank approach o deconsrucng FISIM In deconsrucng FISIM on deposs and loans, we frs apply a narrow bank approach o deermnng he reference rae for he depos akng nsuons, and hen apply he general prncple o he SNA s caegores of loan makng nsuons. In so dong, we end up slghly a varance wh he SNA sandard, n ha we effecvely consder reference raes ha are nsuon-specfc, whle he SNA presumes a reference rae ha s he same boh across nsuons and asse ypes. However, lke he SNA, a leas whn nsuon, our reference rae s no asse specfc. The narrow bank sory old below hus produces a calculaon rule for FISIM ha follows he SNA prescrpon reasonably closely. Depos-akng fnancal corporaons Lke any depos-akng bank, a narrow bank akes deposs and servces deposors, coverng he cos of hese servces by he ncome earned on he deposed funds. A narrow bank, however, does no produce cred servces. We consder frs a smple narrow bank n whch deposors pu her money ogeher no a scheme ha buys rsk free secures and receves neres. I hus assumes no rsk on he asse sde of s porfolo, nvesng deposed funds n lqud, rsk free secures mached n maury o he expeced maury profle of s deposs. The value of he servces of he smple narrow bank o deposors s he opporuny cos of funds nvesed n he rsk free secures less he neres pad deposors. The opporuny cos of deposors funds s dencally her earnngs on he rsk-free porfolo of secures. By mplcaon, he reference rae s hus he depos maury-weghed average rsk free secury rae. Deposors are pad

11 11 all neres less he cos of operang he scheme. FISIM s he dfference beween he opporuny cos of nvesed funds and he amoun pad ou o deposors, whch equals he cos of operang he scheme. The value of cred servces s, by defnon, dencally zero snce here are no loans. Now consder a narrow bank ha employs he capal of s owners, who may be dsnc from deposors, o generae neres ncome on rsk free secures. In hs case, he owners opporuny cos of funds s he rsk free rae approprae o he owners me horzon for parcpang n he busness. The opporuny cos of he owners equy hus would be he neres rae on a rsk free secury whose maury maches he owners expeced duraon of parcpaon n he enerprse. The reference rae hus s he depos and equy mauryweghed average rsk-free secury rae. We expec owners parcpaon duraons o be longer on he average han hose of deposors. When long duraon secures yeld more han shor duraon secures (he usual upward slopng erm srucure of neres raes), he reference rae for a narrow bank wh owners capal hus wll be hgher han for he smple narrow bank. We now move ou of he narrow bank case o consder depos-akng corporaons ha also make loans. Ths covers a broad specrum of bankng nsuons. The SNA recognzes wo ypes of depos-akng fnancal corporaons: The cenral bank Depos-akng corporaons The SNA denfes a hrd caegory of nsuons ha also provde deposor servces: Money marke funds (MMFs). We can vew any depos-akng nsuon as conanng whn a narrow bank, an esablshmen specalzed n deposor servces only. Indeed, boh he cenral bank (usually) and MMFs are essenally narrow banks. 8 We can fnd he value of cred servces on loans as he dfference beween neres earnngs on loans and he reference rae derved from a 8 Cenral banks usually carry governmen secures on he asse sde of her balance shees and commercal bank deposs, currency, and equy on he lably sde. In he exraordnary crcumsances of he recen fnancal crss, he asse sde has ncluded a rsker porfolo of asses such as commercal paper and assebacked secures. Ths exenson n he scope of cenral bank asses s expeced o be unwound as he crss s brough under conrol. Money marke funds closely follow he same paern as he normal cenral bank, carryng shor erm, lqud, low-rsk, usually governmen-ssued secures on he asse sde and deposs (un share accouns) on he lably sde.

12 12 narrow bank wh owners capal. The narrow bank approach o he reference rae hus apples o all depos-akng nsuons. Loan-makng fnancal corporaons oher han depos-akng fnancal corporaons Ths group of nsuons comprses he SNA nsuonal subsecors Oher fnancal nermedares, excep nsurance corporaons and penson funds Capve fnancal nsuons and money lenders As a rule, hese nsuons rely on deb and equy lables as a source of funds wh whch hey make loans. 9 They hus ake our narrow bank wh owners equy o he polar case where deposs are zero. Alhough here are no deposs on he lably sde of her balance shees, deermnng he reference rae for hese nsuons follows he same general prncple as wh a narrow bank wh owners capal. The reference rae would be deermned he same as for he equy componen of he narrow bank s lables, as he owners duraon-weghed average neres rae on rsk-free secures. Ths cerany rae of reurn would leave owners ndfferen beween shares n hese fnancal nsuons and a porfolo of governmen secures whose maury profle maches her me horzons. Assumng owners have a longer me horzon han deposors, and he erm srucure of neres raes s upward slopng, he reference rae for hese non-deposory loan-makng nsuons wll be hgher (or a leas as hgh) as ha for nsuons ha boh make loans and ake deposs. The value of cred servces provded o borrowers (cred FISIM) wll be he dfference beween neres earned on loans and he reference rae of he narrow bank wh owners capal. 10 C. FISIM n he aggregae If we were o add up depos FISIM oupu values across deposory corporaons, he resul would be he dfference beween aggregae neres a he ndusry reference rae (he lably 9 In fac, all fnancal nsuons may oban funds no only from equy, bu also from loans, secures ssues, and oher fnancal nsrumens. 10 As noed n he Inroducon, rsk bearng s among he prmary facor servces he owners provde o any enerprse, ncludng fnancal corporaons. The dfference beween owners and credors reurn on nvesed capal and her rsk-free opporuny cos of money remuneraes hem for hs prmary servce.

13 13 maury-weghed rsk free neres rae for all deposory corporaons 11 ) and he neres pad o deposors. The aggregae reference rae hus reflecs he maury srucure of deposs (among oher lables), lke he SNA. Unlke he SNA, however, also reflecs he srucure of he owners me horzons, bu does no reflec he maury srucure of loans. In addon, unlke he 2008 SNA, here also s no reflecon of he average srucure of rsk 12 prema n he reference rae. Addng up loan FISIM oupu values across loan-makng corporaons, he resul would be he dfference beween loan neres and rsk-free neres reflecng he maury srucure of he lably sde of he balance shee. III. RISK PREMIA AND INSURANCE Our approach o deermnng he SNA reference rae seems o raonalze 2008 SNA gudance on fnancal servces oupu raher well, a leas nsofar as ncorporaes he erm srucure of deposs. Clearly, however, he owners of a bank ake on a leas wo knds of rsk when hey make loans: he loans generally have a longer maury on average han nonequy (depos) loanable funds, and he loans carry a rsk of defaul. The frs s erm rsk, and he second defaul rsk. Basu, Inklaar, and Wang (2008) argue ha all rsk-prema should be excsed from FISIM calculaons because hey are ncome a fnancal nsuon receves for whch has no provded a servce qud pro quo. We examne hs proposon by frs askng wheher a leas par of he rsk prema pay for rsk managemen servces, or nsurance. Followng, for example, Reece (1992), he prce of an nsurance polcy can be expressed p = α δ + ρβ + ( r ρβ )b where p = prce per nomnal un of nsurance n force α = premum rae per nomnal un of nsurance n force δ = expeced loss per nomnal un of nsurance n force β = rae of addons o echncal reserves per nomnal un of nsurance n force 11 Ths would cover all lables from all SNA nsrumens, noably, deposs, loans owed. We ake rsk free raes for loan lables because he FISIM on loans owed s couned as nermedae consumpon of he debor nsuon and hus mus be refleced n he value of oupu. I would be elmnaed from oupu were we o se he reference rae for equy as he raes on he loans hemselves. 12 Excep o he exen ha he benchmark (lowes rsk) secures deermnng he reference rae carry some mnmal level of rsk, n whch case he reference rae wll reflec he srucure of rsk prema for benchmark secures.

14 14 ρ = reference rae r = rae of reurn on polcy loans b = polcy loans per nomnal un of nsurance n force The SNA calculaes he value of non-lfe nsurance servces accordng o he formula Insurance servces = nsurance premums earned + premum supplemens expeced clams. 13 Insurance premums need lle explanaon. Premum supplemens are he nvesmen ncome earned on he nsurance echncal reserves accumulaed o pay clams when nsured evens occur. Ordnarly, f nsurance s purchased from anoher un, we see as an oupu of he producer and servce charges payable for n he expense (nermedae consumpon) lnes of he ncome saemens of he usng uns. However, when hese servce charges are embedded n he ne selemen flows of fnancal dervaves, he SNA consders a ransacon n fnancal asses (fnancal dervaves). I neverheless s conssen wh he SNA o observe ha hese servce charges mus be covered by oupu even f hey only show up as a componen of operang surplus and value added. The SNA reas he non-servce, non-oupu par of he premum, Expeced clams premum supplemens, as a curren ransfer from he nsured un o he nsurer, and, by mplcaon, no par of he value of oupu of he nsurer, whch he SNA erms ne non-lfe nsurance.. When a fnancal corporaon self nsures, he nernal coss of runnng he nsurance operaon appear n par as nermedae consumpon and compensaon of employees, wh he remander appearng n he corporaon s own operang surplus. Under he 2008 SNA, he owners of he nsured fnancal corporaon effecvely capure he ne non-lfe nsurance hemselves. Thus, all of he premum coss of he producon of nsurance for own use, ncludng ne non-lfe nsurance, mus be covered by FISIM. To llusrae, usng he equaon for cred nsurance servces, he loan rae reference rae equals he nsurance servce value + ne non 13 The 1993 SNA followed he Reece (1992) approach very closely for all nsurance, ncludng subracon of ne ncrease n nsurance reserves. The 2008 SNA follows hs formula for lfe nsurance, bu devaes from for non-lfe nsurance by no subracng ne ncrease n reserves. The would hold under he assumpon ha here s no need o accumulae reserves from perod o perod o cover, e.g., relavely rare, caasrophc evens, ha are acuarly mperfecly ancpaed, and ha ne reserves accumulaon s essenally zero; ha s, ha addons o reserves less clams flucuaes closely around zero.

15 15 lfe nsurance, plus ndrecly measured admnsraon and orgnaon fees. By mplcaon, he ne non-lfe nsurance effecvely appears n he purchasng nsuon s value added and operang surplus, wheher purchases he nsurance (va a fnancal dervave conrac) or produces hose servces self. The sysem hus assgns o depos-akng and loan-makng corporaons he ne non-lfe nsurance, even when he nsurance servce s suppled by anoher un. IV. DECONSTRUCTING FISIM ON DEPOSITS A. Accoun servces These servces comprse check clearng and oher accoun servces o deposors. We argue hey are no he only servces FISIM covers. Depos rsk premum B. Insurance (rsk managemen) servces The conrac beween depos-akng nsuons and her deposors s ha deposors accouns are safe and payable n full whn he provsons of he conrac, eher on demand or a he end of a specfed perod (or f earler, on demand subjec o an early whdrawal penaly). As bankng s no enrely whou rsk n meeng he requremens of hs conracual agreemen wh deposors, deposory nsuons underake nernal rsk mgaon or purchase nsurance agans he possbly funds wll no be avalable o mee deposors whdrawal demands. The nsurer manans a echncal reserve o mee deposors clams n he even of an nsolvency even. The nsurer charges depos-akng nsuons a premum fgured as a percenage of deposs for hs nsurance. Depos nsurance servces Depos nsurance servces are calculaed accordng o he non-lfe formula as premums (generally saed n erms of a rae n bass pons per currency un of nsured deposs), plus nvesmen ncome (premum supplemens) on he echncal reserves (also known as he nsurance fund), less expeced losses (expeced payous o deposors as a resul of bank falures). As hs nsurance servce s whn a recognzed class of oupu n he SNA, depos FISIM clearly mus cover. Depos ne non-lfe nsurance The ne non-lfe nsurance componen of he depos nsurance premum s he quany expeced payous o deposors, plus nvesmen ncome on he nsurance fund (premum supplemens). If he ne non-lfe nsurance premum s no covered by depos FISIM,

16 16 operang surplus wll no nclude he rsk prema nvesors requre o employ her capal n he deposory nsuon versus oher nvesmen opporunes. Recen experence on depos nsurance and depos rsk bearng charges In he Uned Saes depos akng nsuons are requred o provde nsurance from he Federal Depos Insurance Corporaon (FDIC), a publc corporaon, on deposs up o a specfed amoun, currenly $250,000 per accoun. 14 The Depos Insurance Fund (DIF) s he nsurance echncal reserve of he FDIC. As of early 2009 was expeced o be abou 1.25 percen of nsured deposs. The depos nsurance premum vares dependng on he assessed rskness of he nsured deposory nsuon, rangng from 5 bass pons per dollar of nsured deposs for low rsk (caegory I) nsuons o 43 bass pons for hgh rsk (caegory IV) nsuons. On January 1, 2009, FDIC rased assessmens o 12 bass pons for caegory I nsuons o 50 bass pons for caegory IV. On Aprl 1, 2009, agan revsed s assessmen mehodology o nclude creds for unsecured deb and surcharges for secured lables and brokered deposs. The mnmum assessmen became 7-24 bass pons for caegory 1 nsuons up o bass pons for caegory IV nsuons. Afer creds, he average nsurance premum rae durng 2008 was 3.8 bass pons per dollar ( percen) of nsured deposs, bu, as noed, wll have rsen snce hen. These rae ncreases have followed rsng assessmens of he expeced losses from bank falures. As losses have been dynamc n recen monhs whle reserves are accumulang more slowly, and hus nvesmen ncome on reserves slowly ncreasng, he rsk bearng charge as a fracon of FDIC s premum has probably been ncreasng snce lae 2007 or early Ths nowhsandng, depos nsurance servces reman small relave o he spread beween he narrow bank reference rae and he rae pad on deposs. For example, he recenly hsorcally hgh roughly 10 bass pon (a enh of a pon of neres) average depos nsurance premum as a percenage of oal (nsured and unnsured deposs) compares wh depos neres pad as a percenage of oal deposs rangng from.89 percen n he frs quarer of 2001 o.21 percen n he frs quarer of In addon o depos nsurance here s he possbly ha reserve requremens now play some role n provdng lqudy nsurance. Such a role was envsoned when hey were nally mposed bu for he pos-war perod hey have nsead served as an nsrumen of money polcy. However, he recen change n polcy ha provdes for he paymen of neres on such reserves may ncrease he lqudy nsurance dmenson. 15 FDIC, Memorandum from Arhur J. Meron, Drecor of Insurance and Research o he Board of Drecors: Assessmen Raes for 2008, March 13, The memo s avalable from he FDIC webse a hp:// A descrpon of FDIC s rsk classfcaon of fnancal nsuons s avalable on s webse a hp:// 16 In addon o he solvency rsk covered (n par) by depos nsurance, varable rae deposs generae erm rsk. We do no deal wh hs depos-sde aspec of erm rsk here. However, we would esmae he nsurance (connued)

17 17 V. DECONSTRUCTING FISIM ON LOANS A. Accoun servces These servces comprse paymen processng and cred rang, among oher servces o borrowers. We argue hey are no he only servces FISIM covers. Loan rsk premum Term or neres rae rsk premum B. Insurance (rsk managemen) servces Loan makng fnancal nsuons lend n long maures and borrow n shor maures. The average porfolo erm rsk premum s he cos of akng hs relavely long poson on he asse sde. Insuons can manage dfferences n he maury srucures of her asse and lably porfolos usng neres rae swaps. These are dervave conracs ha swap a varable shor erm rae generally a London Inerbank Offer Rae (LIBOR) plus a premum wh a rae fxed for he duraon of he swap conrac. Swaps are creaed beween wo pares whou an up-fron charge o eher; ha s, hey are creaed a he money, meanng ha he presen value of he fxed neres sream s equal o he presen value of he varable neres sream. The premum over he varable componen of he swap s he degree of freedom n hs calculaon. I s se o equalze he presen value of he fxed and varable ncome sreams a he naon of he swap. Ths premum, amorzed over he perod of he swap, would be he perod o perod cos of carryng he neres rsk nsurance provded by he conrac. Once he conrac s n place, of course, can go n he money and assume posve marke value, or ou of he money and assume negave marke value, dependng on nvesor vews abou he fuure developmen of he erm srucure of neres raes. 17 Term or neres rae rsk nsurance servces In he case of swaps, he prncpal rsk s counerpary rsk: he even ha one counerpary does no make good on hs oblgaons under he conrac. Hsorcally, counerpary rsk on value of mgaon acves agans hs rsk n he same way as we do for loans below, usng nformaon on neres rae swap prema. 17 Many neres rae swaps also nclude explc servce charges by guaranors of one or boh counerpares, whch would be par of he overall neres rsk nsurance charge bul no he swap arrangemen..

18 18 neres rae swaps has been very low, so we assume counerpary rsk s essenally nl. 18 As a resul, here also are no apprecable echncal reserves held agans counerpary defaul evens. The mplcaon s ha he neres rae rsk nsurance servce charge on swaps s smply he amorzed premum over he varable componen of he swap. In general, we expec ha he longer he agreed duraon of he swap conrac, he hgher he premum, followng he normal upward slope of he maury profle of neres raes. Hence, he dfference beween he rsk free neres cos of he maury profle of he nsuon s lables (deposs) and he maury weghed rsk free neres ncome poenal of he nsuons asses (loans) can be seen as coverng he cos of erm rsk nsurance servces. Term or neres rae ne non-lfe nsurance Because he nsurance servce charge uses he enre premum, he ne non-lfe nsurance on neres rae swaps s nl. Recen experence on erm rsk nsurance (neres rae swaps) Fgure 1 shows he recen hsory of neres rae swap prces. Ineres rae swap premums (defned as he dfference beween he raes on he fxed and varable componens of he swap ) have broadly declned from 7 percen n 2000 o less han 4 percen n The char shows ha shor erm swaps generally le below long erm swaps, bu ha durng perods of moneary ghenng, amorzed premums on shor erm swaps rse o mee premums on long erm swaps as he erm srucure of raes flaens. Daa from he US regulaory auhores pus loan resdual maury a around 12 years, so we chose he 12 year neres rae swap premum as he cos of erm rsk nsurance We noe below, however, ha he same s no rue of counerpary rsk n he cred defaul swap marke. There has been ncreasng pressure from polcy makers n he wake of he fnancal crss o push dervave acvy ono cenral counerpares, and he Basel Commee for Bankng Supervson s proposng o srenghen bank capal requremens for counerpary rsk exposures arsng from dervaves, repos, and secures fnancng acves. 19 FDIC, Call Repors.. Our mplc assumpon s ha shor maures predomnae on he he lably sde of he balance shee. To he exen ha we underesmae maures on he lably sde, we wll oversae he value of neres or erm rsk nsurance ha mus be covered by loan raes on he asse sde. Swaps also may be used o conrol he neres rsk of varable rae deposs on he lably sde of he balance shee. Our focus here s on loans, so we pass over hs addonal erm nsurance componen ha aggregae FISIM mus cover.

19 19

20 20 Defaul rsk premum The defaul rsk premum on a loan s equvalen o an nsurance premum on a conrac ha remburses he loan makng nsuon for losses n repaymen of prncpal (defaul rsk). There are, n fac, such conracs, called cred defaul swaps (CDSs), ha nsuons can purchase or negoae on he fnancal markes. The conrac pays off f a loss-makng cred even occurs n an amoun equal o he dfference beween he conrac prncpal of he loan and he amoun he nsuon recovers from he borrower. The so-called CDS spread s he one perod premum rae on he nsured amoun ha mus be pad o nsure he benefcary of he CDS agans a defaul even. Defaul rsk nsurance servces Agan, he value of he nsurance servce s he premum, plus he premum supplemen (he rae of nvesmen ncome earned on he echncal reserves ha he ssuer of he defaul nsurance holds agans acual losses from nsured cred evens), less he expeced loss. Defaul ne non-lfe nsurance As before, he remander of he premum, gven by he dfference beween he nsurer s expeced losses from defaul and s nvesmen ncome on echncal reserves for CDS conracs, s ne non-lfe nsurance. Recen experence on he value defaul rsk nsurance and ne non-lfe nsurance To provde an ndcaon of he value of he premum componen for defaul nsurance servce oupu, Fgure 2 shows he recen hsory of he defaul nsurance premum measured as quarerly average CDS spreads from 2004, when he daa begn n earnes, unl he second quarer of Lookng a he underlyng daly CDS daa, spreads began around 1.5 percen n 2003, fell o abou 0.9 percen n he mddle quarers of 2004, rose no he 1.5 percen range durng , hen began an accelerang rse n lae 2007, peakng a percen n March of 2009, before rereang o he md 20s by Augus A poenally serous cavea o nerpreng hese defaul nsurance servce esmaes from lae 2007 forward s ha CDS prema rose n large measure based on nvesors belef ha loan defaul probables had ncreased by a large enough amoun (e.g., wh a focus on sub-prme morgages) ha clams on CDS conracs would clmb o levels oo grea o be honored by her ssuers. I hus reflecs a perceved ncrease n he counerpary rsk dosyncrac o CDS nsrumens as much as, or possbly much more han, he ncrease n loan defaul rsk by 20 Source: MarkI,

21 Fgure 2. Defaul Insurance Premum (CDS) and Defaul nsurance Servce Charge CDS premum Defaul nsurance servce charge q1 2001q3 2002q1 2002q3 2003q1 2003q3 2004q1 2004q3 2005q1 2005q3 2006q1 2006q3 2007q1 2007q3 2008q1 2008q3 2009q1 Daa sources: MarkI, CDS prema; Federal Depos Insurance Corporaon, provsons on loan losses, loans, regulaory reserves; US Deparmen of he Treasury, governmen bond raes.

22 22 self. Hence, here lkely was a dvergence beween he cos of marke defaul nsurance and he value of own producon of loan defaul rsk managemen servces by he ssung defaul (CDS) premum raes seen n as he CDS marke began o sablze n 2009, he premum s almos enrely he defaul nsurance servce charge, wh ne non-lfe nsurance consung a very small par of he premum. C. How Do Rsk Managemen Servces Compare wh Loan neres? We argue ha s crcal o use a curren reference rae and he curren marke or far value of loans n calculang loan FISIM, regardless of he 2008 SNA prescrpon ha loans appear on he balance shee a conrac value. Were loans a conac value used o comple FISIM, he reference rae n any gven perod for a loan porfolo would have o be a weghed average of marke reference raes prevalng a he varous loan orgnaon daes n he loan porfolo. Asde from beng daa nensve and complex, hs approach does no reflec curren reurns and valuaons ha should drve he value of servces generaed n he curren perod. Hence, we wre cred FISIM as Loan neres recevable (he reference rae he marke/far value of loans). To esmae he marke value of he loan porfolo, for morgages we obaned raes on morgage backed secures by maury, for commercal and ndusral loans raes on commercal paper, 21 and for raes on consumer loans cred card raes. 22 We esmaed he marke value of morgages, commercal and ndusral, and consumer loans, adjusng he book values of hese loan asses usng hese marke neres raes and he average book reurns on he same asses. We esmae he marke loan rae as a maury weghed average of raes on morgage backed secures for loans secured by real esae and as he average rae on consumer cred for loans no backed by real esae, where he weghs are a esmaed marke values of hese loan classes In he US, commercal paper has an orgnal maury of no more han 270 days, so usng hs an esmaor undersaes he marke rae on loans and oversaes he esmaed marke value of loans, a sgnfcan fracon of whch wll have a longer resdual maury and hgher yeld han secures of 9 monhs or less. 22 Raes on cred cards oversae he average marke rae on consumer loans and undersae he marke value of hese loans. In addon o unsecured cred card deb, much of he remanng non-morgage consumer cred s secured by durable goods and carres lower neres raes. 23 Ideally, we would lke o facor n marke raes earned on commercal loans, bu hese daa were no readly avalable for hs analyss.

23 23 Havng esmaed he defaul rsk and erm rsk componens of he nsurance acves of US banks, n Table 2 we examne how he oal of hese wo componens compares wh he average marke loan rae. The sum of hese wo componens was greaer han or largely covered he dfference beween he conrac neres rae on loans and he nerbank rae hroughou he perod, srongly suggesng ha FISIM calculaed as recommended n he 1993 SNA and sanconed by he 2008 SNA does no undersae he nomnal value of cred servces oupu. Table 2. Componens of Loan FISIM Average esmaed marke rae on loans Average conrac neres rae on loans Dff. n loan raes Toal of defaul nsurance and erm nsurance Inerbank (Federal Funds) rae Toal of nerbank erm nsurance, and defaul nsurance raes Marke loan rae less nerbank, erm and defaul Average Conrac rae less nerbank, erm and defaul 2004q q q q q q q q q q q q q q q q q q q q

24 24 One can see from he Table 1 ha here s a 2-4% dfference beween marke loan raes and average conrac raes. 24 The las wo columns llusrae ha he dfference s especally mporan unl 2006 because of he dfference n sgn beween he rsk managemen adjused user cos, whch can be aken as he loan rae less he sum of he reference rae, he defaul premum and he erm premum. The negave sgn ndcaes ha he cos of purchasng nsurance o manage rsk exceeded he user cos of money on loans (loan FISIM). Ths would nduce banks o ncrease her self-nsurance and/or o raon cred o hgher qualy borrowers. The laer s ndcaed by he repored accumulaon of reserves by banks and he reduced demand for loans by corporaons as evdenced by her accumulaon of cash. Fgure 3 llusraes he rend of he raes. The resdual of he loan rae and he rsk servces componens of oupu covers he reference rae or cos of money and he value of addonal servces besdes rsk managemen. Comparng he sum of hese servce componens wh he loan rae provdes an esmae of he value of cred servces (gross) oupu as he sum of a leas wo of s componens: erm/neres rae and defaul rsk nsurance. Term nsurance s a sgnfcan fracon of he marke loan rae hroughou he perod, and, he crss perod begnnng n he las quarer of 2007 nowhsandng, he nerbank rae plus erm and defaul nsurance servces on US bank loan porfolos has been a leas half of esmaed marke loan neres. Comparng he esmaed marke loan rae wh he sum of he nerbank rae and he nsurance premums for erm and defaul rsk ells an neresng sory before and durng he fnancal crss perod. We see ha banks were operang wh a comforable margn o cover accoun servcng coss durng , bu he margn narrowed begnnng wh Fed ghenng n he hrd quarer of By he frs quarer of 2006, narrowng of he margn slowed. The Fed haled he rse n he nerbank rae n he hrd quarer of 2006 a a level equang our esmaed marke loan rae wh he sum of he erm premum plus he Federal Funds rae n he hrd quarer, despe a slowly growng nsurance premum on defaul rsk. Crcumsances were on he mend by he hrd quarer of 2007, when he sum of he cos of money and rsk nsurance, ncludng defaul rsk, fell back o abou he level of he esmaed marke loan rae. However, by he fourh quarer of 2007, marke assessmens of defaul rsk began a seady upward clmb as he dmensons of he sub-prme morgage problem became evden, o levels far n excess of marke reurns on loans. As noed, however, hs was 24 The 2008 SNA, as earler versons of he SNA, books loans a conrac raher han marke value. All oher fnancal asses are booked a marke value.

25 *Daa sources: Federal Depos Insurance Corporaon, loan neres earned by, and book value of loans of US banks; Fredde Mac Prmary Morgage Survey, raes on morgage loans, Federal Reserve Board, raes on Federal Funds, morgages, cred card deb (consumer cred), and commercal paper; MarkI, cred defaul swap amorzed prema; DaaSream, neres rae swap amorzed prema. 25