A Black Swan in the Money Market 1. John B. Taylor Stanford University. John C. Williams Federal Reserve Bank of San Francisco.

Transcription

1 A Black Swan in he Money Marke 1 John B. Taylor Sanford Universiy John C. Williams Federal Reserve Bank of San Francisco February 21, 2008 On Thursday, Augus 9, 2007 raders in New York, London, and oher financial ceners around he world suddenly faced a dramaic change in condiions in he money markes where hey buy and sell shor-erm securiies. The ineres rae on overnigh loans beween banks he effecive federal funds rae jumped o unusually high levels compared wih he Fed s arge for he federal funds rae. So did he rae on iner-bank erm loans wih mauriies of a few weeks or more, even hough no change in he Fed s arge ineres rae was conemplaed. Many raders, bankers, and cenral bankers found hese developmens srange and surprising afer many years of comparaive calm. The urmoil did no disappear he nex day. The overnigh ineres rae whipsawed sharply down on Friday as he New York Fed pumped liquidiy ino he marke, wih he rae overshooing he arge on he down side by a large margin. Even more worrisome was ha erm iner-bank raes, hose for loans lasing a monh o several monhs, moved up furher on Friday despie he increase in liquidiy provided by cenral banks. Raes on erm lending, such as he Libor one- and hree-monh raes, seemed o have become disconneced from he overnigh rae and hereby from he Fed s arge for ineres raes. 1 We hank John Cogan, Darrel Duffie, Alan Greenspan, Jamie Paerson, Seve Malekian, and Josie Smih for helpful commens. The views expressed in his paper are solely hose of he auhors and should no be inerpreed as reflecing he views of he managemen of he Federal Reserve Bank of San Francisco or he Board of Governors of he Federal Reserve Sysem. 1

2 I was as if banks suddenly demanded more liquidiy or had grown relucan o lend o each oher, perhaps because of fears abou he locaion of newly disclosed losses on subprime morgages. Perhaps some even hough, incorrecly in rerospec, ha he Fed was going o increase he overnigh ineres rae in he near fuure and herefore bid up erm raes immediaely. As we now know, ha Thursday and Friday of Augus 2007 urned ou o be jus he sar of a remarkably long episode of umul in global money markes. The sress has spread beyond hese markes and is by no means over. From he vanage poin of more han a half year laer, he episode looks even more unusual han i did a he sar, and perhaps qualifies as one of hose highly unusual black swan evens ha Taleb (2007) has recenly wrien abou. This episode raises imporan quesions for moneary heory and policy. A a minimum, he sharp changes in risk premia and liquidiy premia provide new daa o sress es our heories of he erm srucure of ineres raes. Moreover, he money marke represens he firs sage of he moneary ransmission channel, where moneary policy acions firs come in conac wih he res of he financial sysem and wih he enire economy. A poorly funcioning money marke jeopardizes he effeciveness of moneary policy. Term money marke raes, such as 3-monh Libor, affec he raes on loans and securiies from home morgages o business loans, so a large spread beween Libor and Fed funds rae can iself inerfere wih how moneary policy affecs he economy. For hese reasons, he Federal Reserve has ried o find ways o reduce he spread beween erm fed funds and he overnigh rae, including lowering he cos of borrowing by banks a he discoun window, direcly encouraging banks o borrow from 2

3 he window, and, mos recenly, inroducing a new way for banks and oher financial insiuions o borrow from he Fed hough he new Term Aucion Faciliy (TAF). We have several aims in analyzing his episode: firs, o documen he unusual developmens in he money markes; second, o assess various explanaions for hese developmens; and hird, o evaluae he impac of various cenral bank policy measures. Though he episode is no over, we feel ha enough ime has passed and enough observaions have accumulaed o draw several conclusions ha are of research ineres and may be useful o policy makers going forward. 1. The Augus 9 Break Poin: Targe, Effecive, and Term Fed Funds Figure 1 focuses on hree money marke ineres raes which nicely illusrae he changes in marke condiions in Augus 2007 (1) he arge for he federal funds ineres rae as se by he Federal Open Marke Commiee, (2) he daily effecive overnigh federal funds rae in he marke, and (3) he ineres rae on 3-monh Libor. The Libor ineres rae in he London iner-bank marke in dollars is essenially he same as he ineres rae on erm fed funds for comparable mauriies, so we focus on he former in his sudy. (Nohing maerial would change if we focused on erm fed funds direcly.) Firs, observe in Figure 1 ha he volailiy of he effecive federal funds rae (he average rae a which overnigh fed funds acually ransac) relaive o he arge increased afer Augus 9. During he period before Augus, he sandard deviaion of he difference beween he effecive funds rae and he arge was only 3 basis poins. Since Augus 9 he sandard deviaion has been 21 basis poins. Noe ha he seadiness of he federal funds rae a 5.25 percen may be one of he reasons for he relaively small 3

4 misses in he earlier period, bu if you include he years back o 2004 he volailiy is 6 basis poins, sill much less han 21 basis poins. There have been oher periods where he effecive funds rae was more volaile, paricularly before he Fed became more ransparen abou is ineres rae seing. See Taylor (2001) for a model ha focuses on effecive fed funds rae volailiy. Second, and his is he main focus of our paper, observe how he spread beween 3-monh Libor and he Fed s overnigh federal funds rae arge increased dramaically saring in Augus and flucuaed erraically afer ha. During he year before Augus 2007 he 3-monh Libor spread above arge federal funds averaged only 11 basis wih a sandard deviaion of only 1 basis poin a period of very low volailiy. Similar changes in spreads beween erm raes and overnigh raes are apparen for oher Libor mauriies and for several oher counries, as we documen below. Percen Targe Federal Funds Effecuve Federal Funds LIBOR 3-monh Sep Nov Jan Mar May Jul Sep Nov Jan Figure 1. Key money marke raes from Sepember 2006 o February

5 2. Poenial Explanaions Ever since he urmoil began, raders, bankers, economiss, and many ohers have offered explanaions for he increase in he Libor spread. We hink i is useful o caegorize he many explanaions ino several ypes. Firs, and perhaps he mos commonly menioned, is counerpary risk, which simply means banks became more relucan o lend o oher banks because of he percepion ha he risk of defaul on he loan had increased and/or he marke price of aking on such risk had risen. Recall ha iner-bank lending in he Libor marke or erm fed funds is unsecured. Of course, his explanaion has he virue of reflecing he widely-repored realiy ha many banks were wriing down securiies ha had eiher been downgraded or were backed by sub-prime morgages ha were becoming delinquen or going ino foreclosure as housing prices sopped increasing and began o fall. Clearly he coninuing decline in housing prices and he slowing economy could easily raise he chances of furher bad debs on he banks balance shees. Moreover, he realizaion of he risks in derivaive securiies based on sub-prime morgages riggered doubs abou many oher aspecs of he derivaive marke including he abiliy of credi defaul insurers o mee heir obligaions and he size and naure of he likely resrucuring of he off balance shee operaions know as SIVs Anoher explanaion, which migh be called liquidiy risk, is ha raders a one bank are relucan o expose he rader s bank s funds during a period of ime where hose funds migh be needed o cover he bank s own shorfalls. Effecively, he rader may no 5

6 be given as much balance shee o inves, which is perceived as a shorage of liquidiy o he rader. A hird and closely relaed explanaion was ofen heard during he period of November and January, I was ha banks needed liquidiy o make sure ha heir own balance shees looked respecable in end-of-year financial repors, especially under he sress and scruiny ha many banks had been under. The fourh explanaion is due o expecaions of fuure ineres rae changes. Excep for he very beginning of he urmoil period his explanaion would end o bring he spreads down because of expecaions of fuure ineres rae decline due o policy easing. Neverheless i is essenial o ake accoun of his facor when assessing he oher facors ha could be moving he spread around. For example, if you look closely a Figure 1 you see ha spread beween Libor and he fed funds arge comes down before cus in he federal funds rae. Indeed, oward he end of our sample in mid February, he spread had narrowed significanly, bu his could be due o expecaions of fuure ineres rae cus. We herefore conrol for expecaions of fuure ineres raes in he analysis ha follows. 3. A Model In order o disinguish beween hese various explanaions we need a model of money marke ineres raes hrough which we can inerpre he risk, liquidiy, and expecaions facors ha we have argued are imporan. I is essenial o ake ou pure expecaions effecs, which always creae differences beween longer erm ineres raes 6

7 and overnigh fed funds. Recall ha Libor is a erm rae (3 monh in Figure 1) and fed funds are one-day mauriy. Early models of he money marke used for moneary policy developed in he1970s and 1980s (see Anderson and Rasche (1982) for a review) are no sufficien for his purpose because hey neiher accoun for forward-looking expecaions nor risk premia. More recen finance models used by Ang and Piazzesi (2003) and McGough, Rudebusch and Williams (2005) are more useful for his purpose. Moreover he earlier models used esimaed demand funcions for securiies, an approach ha is no possible o implemen in he curren siuaion because available daa is in he form of prices (in he form of ineres raes), raher han quaniies. i ( n) s a ( n) (n) Our model focuses on hree ineres raes as defined below: = libor rae wih mauriy n (wih n = 1defined o be he overnigh federal funds rae) = OIS wih mauruiy n (wih n = 1also he overnigh federal funds rae; ha is s = acceped bid on he erm acion faciliy (TAF) ( n around 30 days) (1) = i ( 1 ) ) The Overnigh Indexed Swap (OIS) rae is closely conneced o he average overnigh ineres rae expeced o prevail over he nex n days. An OIS is srucured as follows: a mauriy, he paries exchange he difference beween he ineres ha would be accrued from repeaedly rolling over an invesmen in he overnigh marke and he ineres ha would be accrued a he agreed OIS fixed rae. The TAF is described in deail below. Following he lieraure on arbirage-free pricing of bonds, we wrie down erm srucure relaions for he Libor (or fed funds) erm srucure ineres raes. Le denoe he price of a zero-coupon loan wih n periods unil mauriy. Equaion 1 relaes he yield on he loan, (n) i, o is price. The prices of zero-coupon loans follow he (n) P 7

8 recursion given in equaion 2, where m+ 1 denoes he pricing kernel. As in Ang and Piazzesi (2003), we assume he pricing kernel akes he form shown in equaion 3 and he marke price of risk, λ, akes he linear form shown in equaion 4, where x is a vecor of variables ha affec he price of risk. (1) i ( n) = n 1 log( P ( n) ) (2) P ( n+ 1) = E [ m + 1 P ( n) + 1 ] (3) m + 1 = exp( i (1) 2 0.5λ λ ε + 1 ) (4) λ = γ γ x 0 1 Similar equaions can be wrien down for he OIS and he TAF raes. In conras o Libor loans, OIS ransacions involve very lile counerpary risk as no money changes hands unil he mauriy dae. The only poenial loss in case of defaul by he counerpary is he difference beween he wo ineres raes on which he OIS is based. There exiss some ineres rae risk reflecing uncerainy regarding he fuure pah of ineres raes. However, given he shor mauriies of up o hree monhs ha we analyze in his paper, he marke price of ineres rae risk is likely very small, and as an approximaion we assume ha i is zero. Loans from he TAF are collaeralized and herefore also carry relaively small risk. We herefore assume ha he marke price of risk associaed wih TAF loans is likewise zero. Taken ogeher, his assumpion of a zero marke price of risk for OIS and TAF raes implies ha as par of he null hypohesis of an absence of liquidiy effecs in he 8

9 pricing of he various loans, we have: a = s. Moreover, absen liquidiy effecs, we ( n) ( n) would no expec he λ i for he iner-bank raes o be influenced by he TAF. Under hese assumpions, he OIS rae equals he average of he overnigh nigh ineres raes expeced unil mauriy. By subracing he appropriae OIS rae from he erm Libor yield, we are able o cleanse expecaions effecs from he Libor yield. Under our null hypohesis of no liquidiy effecs, he resuling difference in raes, i s, ( n) ( n) reflecs only he pricing of risk associaed wih Libor lending. Thus, in he nex secion, we use his difference in yields as a measure he effecs of risk on yields. We will use several differen measures of counerpary risk as explanaory variables in he price of risk, as explained below. 4. Focusing on he Libor OIS spread. Figure 2 plos he spread beween Libor and OIS during he same period as in Figure 1. I pains quie a differen picure of he spread, and shows he value of removing expecaions of fuure ineres raes in analyzing erm spreads. For example, looking a Figure 1 you migh hink he spread reurned o normal by mid February. However, examinaion of Figure 2 shows ha he spread is sill quie large. In his char and in he res of our analysis we focus on 3 monh Libor; similar resuls are found by looking a oher mauriies such as one-monh Libor. Figure 2 illusraes clearly how he spread beween Libor and OIS jumped on Augus 9 h. From December 4, 2001 he day when our OIS 3-monh daa begin hrough 8 Augus 2007, he spread averaged 11 basis poins wih a sandard deviaion of 3.6 basis poins. I jumped by 25 basis poins above his average o 34 basis poins on 9

10 Augus 9 h, and since hen i has flucuaed widely beween a minimum of 30 basis poins and a maximum of 106 basis poins; i has averaged 65 basis poins. On February 15, i was 53 basis poins slighly less han he average since Augus 9, 2007 clearly no a reurn o a normal level. The peak was on December 6, 2007, and here was a big downward movemen on December 12-14, 2007, and a more gradual decline began unil anoher big jump down occurred on January 14-15, Percen monh LIBOR OIS Spread Sep Nov Jan Mar May Jul Sep Nov Jan Figure 2. Taking ou he pure expecaions effecs and leaving in he risk and liquidiy effecs. Looking a spreads going back o December 2001 illusraes jus how unusual his episode has been. Figure 3 plos he same daa as in Figure 2, bu saring in December As menioned above, he spread on Augus 9 was 25 basis poins above he pre- Augus 9, 2007 average. Tha is 7 imes he sandard deviaion before Augus 9 more 10

11 han a 6-sigma even. The mean since hen is 16 sandard deviaions above he old mean, which under normaliy would have been an exraordinarily improbable even. Percen monh libor ois spread Figure 3. A Black Swan in he Money Marke? Is i possible o ge a longer perspecive on he recen episodes of risk/liquidiy premia on Libor? Thus far our comparison only goes back o lae 2001, he ime ha OIS daa sared being colleced. While he large spread observed since Augus is unusual compared wih his recen period, he longer hisory of hese yields should also be examined. Figure 4 is useful for his purpose. I shows he spread beween unsecured iner-bank lending (Libor) and secured iner-bank Repo (Repurchase Agreemens backed by Treasury securiies) lending of he same mauriy, in his case hree monhs. Noe ha his measure of risk spreads using Repo raes o conrol for expecaions effecs shows he 11

12 same large increase in spreads afer Augus 9. This measure of risk spreads experienced some episodes of shor-lived spikes in he 1990s, indicaing ha he curren episode is no as highly improbable as implied by he more recen evidence based on Libor-OIS spreads. Noneheless, hese pas episodes were no nearly as large or persisen as ha experienced in he pas six monhs. Percen monh LIBOR less 3 monh Repos Figure 4. A Longer Perspecive 5. Overnigh Funds Volailiy: Counerpar Risk or Increased Tolerance o Misses Thus far we have shown how imporan i is o ake ou expecaions effecs in order o assess he increase in risk and liquidiy premia in he inerbank marke. I is also possible o focus direcly in he increase in volailiy of he effecive funds rae relaive o 12

13 is arge as se by he FOMC. Figure 5 shows he difference beween he effecive fed funds rae and he arge fed funds rae. There are several possible explanaions for he increased volailiy (or misses of he effecive rae from he arge). One is he same counerpary risk ha is offered as an explanaion for he spread seen in he erm lending marke. Fed funds rades are largely bilaeral. Hence raes can differ from rade o rade even a he same poin in ime. If raders are more circumspec abou some borrowers han ohers hen his will show up in increased volailiy of he effecive rae, which is esimaed from he bilaeral rades. Tha volailiy has increased in he overnigh marke is herefore corroboraing evidence ha couner pary risk may be par of he explanaion for he increased spread in he erm marke. Anoher explanaion, however, is ha he underlying volailiy in inraday rading in he fed funds marke and he New York Fed s rading desk has aced o preven he rae from spiking on he up side. Indeed here is a noiceable downward bias in he misses in 13

14 he pas six monhs. Percen Effecive Federal Funds less Targe Figure 5. Increased Volailiy in he Overnigh Federal Funds Marke 6. Measures and Indicaors of Counerpary risk In his secion we consider a range of possible indicaors of counerpary risk. To he exen ha hese are imed wih he black swan even documened in Figure 2, hey may offer evidence ha such sources of risk, raher han more general liquidiy concerns, have been he main reason for he increased spread in he Libor markes. Asse Backed Commercial Paper versus Dealer Placed Commercial Paper Anoher marke ha has been under exreme sress during his period is he marke ha grew as a mechanism for financing he purchase of home morgages in he process of assembling hem ino various derivaive securiies. Because he commercial 14

15 paper was backed by hese morgages or by he morgage pools, hey are called assebacked commercial paper. They are a poenial measure of he counerpary risk in commercial banks because banks held his paper eiher direcly or indirecly hrough heir SIV operaions. Figure 6 shows he spread beween asse-backed commercial paper and dealerplaced commercial paper, which excludes he more risky asse-backed issues, leer-ofcredi issues, and direc issues from firms. Clearly here was an increase in he spread abou he same ime as he Libor spreads increased. Ineresingly he paerns of decline and he ups and downs also have similariies. To he exen ha his is a good indicaor of couner pary risk, his iming lends suppor for he counerpary risk explanaion. 15

16 Percen Asse Backed - Dealer Placed Commercial Paper Spread 30-day (op ier) JanFebMarAprMayJun Jul AugSepOc NovDec JanFeb Figure 6. Asse Backed Commercial Paper Spreads Increased abou he Same Time as Libor Spreads Credi Defaul Swaps Anoher measure of couner pary risk is he probabiliy ha banks migh defaul on heir deb. These probabiliies can be assessed using he premiums on credi defaul swaps ha are like insurance policies for corporae bonds. The buyer of a credi defaul swap (CDS) pays a periodic fee o a seller in exchange for he promise of a paymen, in he even of bankrupcy or defaul, of he difference beween he par value and he marke value of he corporae bond. Figure 7 shows he raes on five-year CDS for several major financial insiuions in recen years. Noe he increase saring in July of Figure 8 focuses on he hree large commercial banks. Unlike he asse backed commercial paper 16

17 spread, here is no evidence of a decline in risk his year a he ime ha he Libor spreads declined. Basis Poins Wells Fargo Bear Searns Goldman Sachs Bank of America Ciigroup HSBC Merrill Lynch 0 Jul Aug Sep Oc Nov Dec Jan Feb Figure 7. Risk a Banks as measured by CDS raes increased in he summer of

18 Basis Poins Credi Defaul Swaps - 5-year Wells Fargo Bank of America Ciigroup Jul Aug Sep Oc Nov Dec Jan Feb Figure 8. Risks as measure by CDS a hree major banks coninued increasing Developmens in oher Counries Ye anoher way o ge a he risk in he iner-bank marke is o look a Libor spreads in oher currencies. Euro Libor and Pound Serling Libor Figure 9 shows he Libor spreads for loans in Euro and Pound Serling using he same OIS adjusmen mehod we used above o calculae he U.S. dollar Libor spreads in Figure 2. We plo hese oher wo spreads along wih he dollar spread since All hree spreads move closely ogeher, indicaing ha whaever he source of hese spreads, i is affecing money markes for all hree currencies in he same way. 18

19 Percen monh LIBOR OIS Spreads US EU UK Figure 9. Libor spread increased in hree major currencies in Augus 2007 Yen Libor and Tibor. Anoher useful indicaor is a comparison of he Libor rae denominaed in Yen o ha of he Tibor, he rae on iner-bank loans beween Japanese banks in he Tokyo markes. Figure 10 shows he wo raes since he mid 1990s. Noe ha he char shows he Libor yields hemselves, no spreads. Japanese ineres raes have been much lower han ineres raes in he Unied Saes, Europe or he UK. Noneheless, spreads can and do develop beween differen ypes of iner-bank lending and indicae risk facors in he banking secor. Indeed, in he lae 1990s Japanese banks experienced sizable spreads on iner-bank lending comparable o wha is being experienced in New York and London in his recen episode of sress. As explained by Peek and Rosengren (2001) and by Corvig, Low, and Melvin (2004), risks in he banking secor in Tokyo caused ineres raes on iner-bank loans o rise in Tokyo compared wih 19

20 London. In oher words, Tibor raes rose relaive o Libor raes, as shown in Figure 10 and Figure 11, which shows he Tibor-Libor spread for hree-monh loans. Percen Tibor 3-monh Libor 3-monh Figure 10. Paern of Tibor and Libor since 1990s This paern of Tibor-Libor spreads has reversed, wih Tibor raes now lower han corresponding Libor raes. One inerpreaion is ha he demand for liquidiy has no risen as much for Japanese banks as for he major banks in hese oher markes. In our view, a more probable explanaion is ha he risks associaed wih iner-bank loans from American and European banks have increased relaive o hose for loans among Japanese banks. Accordingly, he negaive Japan premium or Japan discoun provides anoher measure of counerpary risk among banks in New York, London, and Frankfur. 20

21 Percen Tibor-Libor 3-monh (Japan Premium) Figure 11. Unlike he Japan premium in he 1990s he Tibor-Libor spread urned negaive fell when Libor spreads increased in he Unied Saes and Europe Swiss Libor. Finally we look a Libor loans denominaed in Swiss Francs. The Swiss Naional bank follows a differen sraegy for moneary operaions han he Federal Reserve, he European Cenral Bank, or he Bank of England. The Swiss Naional Bank arges he hree-monh Libor rae and adjuss he amoun of liquidiy in he overnigh marke o hi is arge. Hence, if here is an increase in he spread beween hree-monh Libor and he overnigh rae, hen he SNB will ake acions o reduce he overnigh rae by providing exra liquidiy o he marke. (See Jordan Kugler (2004)). As a resul, a very differen paern emerges in he overnigh and erm Libor raes. However, he same evidence of risk emerges if one looks a he spread beween overnigh and erm raes. 21

22 These acions can be seen clearly in Figure 12. Wih a arge for 3-monh Libor of 2.75 percen, he overnigh rae declined emporarily while he Libor rae remained seady. Hence, he spread beween Libor and overnigh was realized by a lowering of he overnigh rae. The way his works is nicely illusraed in he period from Augus hrough February of The Swiss Libor rae firs increased a bi in Augus and hen was brough o arge. Par of his iniial increase was anicipaion of an increase in he arge, bu par was evidenly due o an increase in risk. Figure 12. Term Libor spread in Swizerland resuled in a emporary decline in he overnigh rae wih curren operaing procedures a he SNB 22

23 7. The Term Aucion Faciliy In an effor o lower he unusual erm lending spreads documened in Figure 2, he Federal Reserve has aken a number of acions. Firs i lowered he spread beween he discoun rae and he fed funds arge direcly and encouraged more discoun window borrowing. Bu, banks did no increase here borrowing o any large degree. Second, in December 2007, he Federal Reserve esablished a new faciliy called he erm aucion faciliy (TAF) o provide liquidiy direcly financial insiuions a a longer duraion, and hereby drive down he spread on erm lending relaive o overnigh loans. According o he Federal Reserve Board, by injecing erm funds hrough a broader range of counerparies and agains a broader range of collaeral han open marke operaions, his faciliy could help ensure ha liquidiy provisions can be disseminaed efficienly even when he unsecured inerbank markes are under sress (Board of Governors of he Federal Reserve, 2007). The TAF was firs announced on Dec 12. The TAF allows financial insiuions o make bids for erm borrowing from he Fed (wih mauriies ypically of 28 days). So far, five TAF aucions have been held: December 17, December 20, January 14, January 20, and February 11 and more are scheduled o be held. Table 1 provides he key informaion abou each of hese aucions. Figure 13 shows he daes wih verical lines. TAF loans are collaeralized following he procedures used for discoun window borrowing. The Board of Governors ses he aucion amoun and he minimum ineres rae bid, which is se equal o he OIS rae corresponding o he erm of he loan. The ineres rae on he loans is deermined by an aucion and is repored as he sop-ou rae in Table 1. The spread beween he aucion sop-ou raes and he OIS rae a he ime 23

24 bids were aken averaged around 50 basis poins for he firs wo aucions, bu for he subsequen aucions his spread has been beween 2 and 15 basis poins. Table 1. Term Aucions wih TAF Bid Sele Term Am Min Sop-ou Spread Bid/Cover # bidders Day Day (days) ($B) Bid 17-Dec 20-Dec Dec 27-Dec Jan 17-Jan Jan 31-Jan Feb 14-Feb Percen 1.2 Lines show daes of TAF bids monh LIBOR OIS spread 0.0 Jul Aug Sep Oc Nov Dec Jan Feb Figure 13. Timing of he TAF aucions 24

25 In assessing is effecs, i is imporan o noe ha he TAF does no increase he amoun of oal liquidiy in he money markes. Any increase in liquidiy ha comes from banks borrowing from he Fed using he TAF will be offse by open marke sales of securiies by he Fed o keep he oal supply of reserves from falling rapidly. The acions are essenially auomaic in he sense ha he Fed mus sell securiies o keep he federal funds rae on arge. Figure 14 shows ha his is indeed wha has happened under he TAF. The Sysem Open Marke Accoun reduced is ourigh holdings of securiies (ligh blue area) by essenially he same amoun as he TAF (dark blue area). This can also be seen in Figure 15: Noe ha TAF borrowings have increased dramaically only o be compleely offse by a sharp declined in non-borrowed reserves leaving oal bank reserves a he Fed largely unchanged. Source: Federal Reserve Bank of New York, Domesic Open Marke Operaions in 2007, February 2008 Figure 14 TAF did no increase he oal amoun of liquidiy 25

26 Billions of dollars Borrowings (including TAF) Non Borrowed Reserves Jan 07 Apr 07 Jul 07 Oc 07 Jan 08 Figure 15. As TAF borrowings from he Fed go up, non-borrowed reserves decline o offse he increase, keeping oal reserves unchanged 8. Economeric Tess In his secion we endeavor o es wheher he various risk variables can explain he Libor-OIS spread using he measures explored in previous secions. The es is performed wih simple regressions. The null hypoheses are ha (1) he risk facors do no significanly affec he spread and (2) he TAF does no significanly affec he spread. In each regression we use he daily daa presened in he chars and look a he sample period from January 2, 2007 hrough February 15, 2008, a span of ime ha includes boh he marke urmoil period and a comparable period of ime before he urmoil. The dependen variable in each case is he Libor3 OIS3 spread. The independen variables are various indicaors of counerpary risk as described in he ile o each able and a TAF dummy (TAFD) which is one on each of he five TAF bid daes 26

27 and zero elsewhere. There are four ses of regressions corresponding o differen risk measures. For each of he risk measures, we repor OLS regressions as well as regressions correced for firs-order serial correlaion (AR(1)) wih he esimaed serial correlaion coefficien ρ repored.. Table 2 Asse Backed Commercial Paper Spread: OLS: R 2 =.74 Variable Coefficien Sd. Error -Saisic Consan SPREADABCP TAFD AR(1): ρ =.982 Variable Coefficien Sd. Error -Saisic Consan SPREADABCP TAFD Table 3. Credi Defaul Swap on Ciigroup: OLS: R 2 =.59 Variable Coefficien Sd. Error -Saisic Consan CCGI1U TAFD AR(1): ρ =.981 Variable Coefficien Sd. Error -Saisic Consan CCGI1U TAFD

28 Table 4: Credi Defaul Swap on Bank of America: OLS: R 2 =.94 Variable Coefficien Sd. Error -Saisic Consan CBAC1U TAFD AR(1): ρ =.983 Variable Coefficien Sd. Error -Saisic Consan CBAC1U TAFD Table 5: Tibor-Libor Spread: OLS: R 2 =.61 Variable Coefficien Sd. Error -Saisic Consan JAPANPREMIUM TAFD AR(1): ρ =.987 Variable Coefficien Sd. Error -Saisic Consan JAPANPREMIUM TAFD The common heme of all hese resuls is ha (1) one can easily rejec he null hypohesis ha he counerpary risk facors are no significan in he Libor OIS spread and (2) one canno rejec he null hypohesis ha he TAF has no effec. 28

29 9. Conclusion In his paper we documened he unusually large spread beween erm Libor and overnigh ineres raes in he Unied Saes and oher money markes since Augus We also inroduced a financial model o adjus for expecaions effecs and o es for various explanaions ha have been offered o explain his unusual developmen. The model has wo implicaions: firs ha counerpary risk could be a facor in explaining he spread beween he Libor rae and he OIS rae, and second ha he TAF should no have an effec on he spread. Since he TAF does no affec oal liquidiy, expecaions of fuure overnigh raes, or counerpary risk, he model implies ha i will no affec he spread. Our simple economeric ess suppor boh of hose implicaions of our model. 29

30 References Anderson, Richard C. and Rober H. Rasche, Wha Do Money Marke Models Tell Us abou How o Implemen Moneary Policy, Journal of Money Credi and Banking, 1982, Vol. 14, No. 2, Par 2 Ang, Andrew and Monika Piazzesi (2003), A No-Arbirage Vecor Auoregression of Term Srucure Dynamics wih Macroeconomic and Laen Variables, Journal of Moneary Economics (May), 50, 4, Board of Governors of he Federal Reserve, Term Aucion Faciliy FAQs, December, Corvig, Viceniu, Buen Sin Low, and Michael Melvin (2004), A Yen is no a Yen: TIBOR/LIBOR and he Deerminans of he Japan Premium, Journal of Financial and Quaniaive Analysis, 39, 1, Jordan, Thomas J. and Peer Kugler (2004), Implemening Swiss Moneary Policy: Seering he 3M-Libor wih Repo Transacions, Swiss Naional Bank (May 23). McGough, Bruce, Glenn B. Rudebusch, and John C. Williams (2005), Using a Long- Term Ineres Rae as he Moneary Policy Insrumen, Journal of Moneary Economics (July), 52, 5, Peek, Joe and Rosengren, Eric S. (2001), Deerminans of he Japan Premium: Acions Speak Louder han Words, Journal of Inernaional Economics, Vol. 53, pp Taylor, John B. (2001), Expecaions, Open Marke Operaions, and Changes in he Federal Funds Rae, Review, Federal Reserve Bank of S. Louis, Vol. 83, No. 4, July- Augus, pp Taleb, Nassim Nicholas (2007), The Black Swan: The Impac of he Highly Improbable, Random House, New York 30