Liquidity Black Holes: what are they and how are they generated

Transcription

1 : what ae they and how ae they geneated Avinash Pesaud Gesham Pofesso of Commece Managing Diecto and Global Head of Reseach, State Steet This aticle was published in the Singapoe Foeign Exchange Maket Committee Biennial Repot, Thanks go to Natalia Alvaez-Gijalba of State Steet and Stephen Spatt of Sussex Univesity fo statistical assistance.

2 Why does liquidity matte? Financial liquidity mattes. In geneal maket paticipants have a stong pefeence fo tading in the most liquid makets, which patly explains the ecent pessue fo meges and coss-listing ageements between stock makets and the consolidation of electonic bond and cuency exchanges. In a seies of patneships and coss-listings, Euopean exchanges have coalesced aound two goups and the London Stock Exchange is expected to ente into a patneship o mege shotly. The Singapoe and Austalian stock exchanges have enteed into a patneship designed to suppot liquidity. As impotant as the level of liquidity, is its uncetainty. In an age whee thee is an intoleance fo isks that cannot be quantified, investos avoid makets altogethe whee liquidity is uncetain. This is one of the easons why potfolio flows to emeging makets ae concentated in a vey small numbe of lage makets, see chat 1. Thee is some evidence that investo pefeence fo liquidity inceased following the collapse of LTCM in Septembe 1998, a hedge fund whose abitage stategies fatally depended on the maintenance of liquidity. It is of paticula concen, theefoe, that tades fequently talk of episodes in which the liquidity faced by a buye o selle of a financial instument vitually vanishes, eappeaing again a few days o weeks late. Moeove, these liquidity black holes, as I have coined them, ae not confined to the usual suspects such as the Indonesian upiah maket in August 1997 o the South Afican and in Apil 1998, but ae said to occu in the lagest makets. Tades egulaly swap wa stoies about liquidity black holes in the dolla/yen o steling/dolla exchange ate o the US Teasuy maket. Liquidity black holes wee cetainly pesent in the collapse of majo equity makets outside Asia in Chat 1: The popotion of goss coss-bode equity flows to 17 emeging makets accounted fo by just five makets. Pecent Top 5 counties All othe counties Souce: State Steet Bank Page 2 of 9

3 Issues of concept and definition Liquidity and volatility ae elated, but sepaate concepts. A lack of liquidity will in most cases lead to volatility and so the ise in volatility in majo financial makets is suppotive of the tades anecdotes, see chat 2. But it is not poof. Liquidity black holes will lead to highe volatility, but not all instances of highe volatility stem fom a liquidity black hole. The public announcement of new and supising infomation which leads tades to move thei bids significantly to a new level will appea as pice volatility, but eflects a well functioning maket moving to a new equilibium, not an illiquid one. An ill-functioning maket would be one whee that public announcement causes the pice of an instument to fall and that fall then causes some maket paticipants to sell, pushing the pice lowe still, causing moe to sell. Pice falls tiggeing futhe sales. This is what is said to have occued in dolla/yen tading in Septembe 1998, o in the 30 yea US Teasuy maket in Febuay 2001 o in the UK equity maket this yea as UK insues esponded to pice declines by selling equities. The kind of pice move obseved in a feedback loop like this is of a vey diffeent chaacte to the oiginal public announcement effect. In this unstable envionment liquidity to sell (o buy) vitually vanishes only to eappea when the maket stabilises. By a stable maket we mean a maket whee buyes ae bought out by lowe pices and selles by highe pices. This definition opens up a dimension of hoizons. A maket may have many mico-black holes when looked at ove peiods of a few minutes, but is stable when looked at ove peiods of a few days. The longe these peiods last the moe concening they ae. We ae concened with time peiods long enough - say a few days and pehaps as long as a week - fo liquidity black holes to yield significant costs to those who ae dependent on tading on a paticula day, say month o quate end, o long enough to change an investment decision o at least potentially to do so. An inceasingly lage numbe of financial contacts use a maket pice of an instument on a specific date o time on the assumption that the maket pice will eflect undelying valuation and not liquidity. Page 3 of 9

4 Chat 2: The numbe of days in a quate that the aveage pice of the thee lagest equity makets has moved by 2 standad deviations moe than the aveage daily pice move Numbe of days 60 Souce: State Steet Bank 5 yea smoothed aveage Q1 79 Q1 80 Q1 81 Q1 82 Q1 83 Q1 84 Q1 85 Q1 86 Q1 87 Q1 88 Q1 89 Q1 90 Q1 91 Q1 92 Q1 93 Q1 94 Q1 95 Q1 96 Q1 97 Q1 98 Q1 99 Q1 00 Q1 01 Q1 02 Q1 A hypothesis about liquidity black holes and divesity The pesence of liquidity poblems in the lagest of makets suggests that liquidity is not about size, but divesity. In an illiquid maket the same size of sell ode will push the maket down futhe than in a liquid maket. Imagine a maket whee thee is a lage numbe of maket paticipants, using the exact same infomation set, in the exact same way, to tade the exact same financial instuments. When one buys they all do and vice vesa. Maket paticipants would face volatility and illiquidity when they came to buy o sell. This would not be educed by having moe playes, only by inceasing the amount of divesity in thei actions. (Indeed, on these assumptions it is possible to show that the bigge the maket was, the less liquid it would be). Now imagine a maket with just two playes but with opposite objectives o opposite ways of defining value. When one wants to buy the othe wants to sell. This maket is small, but the pice impact of tading would be low and liquidity would be high. At fist sight, ou imaginay maket whee thee is no divesity is fa fom eality. Howeve, only a few changes ae equied to bing us to a eality filled with black holes. Thee ae makets which ae lage in tems of daily tunove but ae dominated by a few instuments. As infomation costs have collapsed we inceasingly shae the same infomation, by and lage, and egulatoy foces ae encouaging moe timely and democatic disclosue of company infomation. We do not stat off using this infomation in the same way, as in ou exteme example, howeve, we may end up doing so as a esult of a combination of uncetainty and the use of common, maket sensitive isk management systems. Page 4 of 9

5 In an uncetain wold, maket paticipants collect in heds. If I feel I don t know whee I should be going but I think othes may know, my best stategy is to follow what othes ae doing. Moeove, if govenments bail out financial fims only when thei combined failue will impact the financial system, but not when an individual failue will leave the system unscathed, thee ae fa moe isks to being wong and alone than being wong and in a cowd. This is why banks have an uncanny habit of making spectacula investment mistakes togethe such as duing the Latin Ameican debt cisis of the ealy 1980s, the popety cash in the UK and Scandinavia in the late 1980s, the Asian financial cisis in the mid-to-late 1990s and the dot.com debacle at the tun of the decade. Maket-sensitive isk management systems equie maket paticipants to educe thei exposue to isk when volatility ises and/o pices decline. In esponse to a negative shock, a hed, using maket sensitive isk management systems will sell the same instuments at the same time and the futhe pice decline will tigge futhe sales aising volatility which will also tigge futhe sales. Maket-sensitive isk management systems ae pevalent and ae becoming moe so. They take many foms. They include sophisticated value-at-isk systems used by banks and isk limits imposed on thei tades o magin calls. Thee ae also solvency tests fo insues o limits on pay-out ates. Regulatos ae ovely fond of maket-sensitive isk systems especially when they focus on the mico institutional aspect of isks to the neglect of the maco-pudential aspects. Makets pone to black holes The foeign exchange maket is one of the lagest makets in tems of daily tunove with an estimated $1.5tn of foeign exchange spot and fowads taded evey day. It is dominated by just thee majo instuments. Dollas, yen and euos ae on one side of ove 80% of foeign exchange tansactions. Just 12 banks account fo 75% of all foeign exchange tading. These lage intenational banks have been encouaged by egulatos to use maket sensitive isk management systems. In shot, FX is a lage maket dominated by few instuments and few playes, using simila infomation sets and decision ules. It should be ife with liquidity black holes despite its enomous size. The bond maket is not quite the exact opposite, but nealy so. In ecent yeas the lagest bond maket, the US Teasuy maket, has shunk and tunove has fallen off. Because US govenment bond yields ae global benchmaks they ae taded and held by a divese set of playes fom banks and insuance companies to pension funds, mutual funds and cental banks. Being benchmaks they ae used fo many diffeent things - investments, picing fo issuance, hedges, safe-havens etc. The lage ole played by the public secto and quasi-public secto who ae not egulated as pivate financial institutions adds consideably to divesity of playes, stategies and decision ules. Bond investos also have a lage numbe of altenatives to choose beyond US Teasuies with divese matuities, size of issue and cedits. Thee ae moe bond instuments than equity instuments. Page 5 of 9

6 If divesity is key, the 10 yea US Teasuy maket - a global bond benchmak - should have less liquidity black holes than the dolla/yen maket. If size was key, howeve, equity makets would have moe liquidity black holes, bonds next and majo cuency makets last. Table 1: Ranks of divesity and tunove Ranks 1 = most, 3 = least Playes Instuments Risk Management Tunove Bonds Equities Cuencies A test: To test fo liquidity black holes and the ill o well functioning of makets we want to investigate the diection of causality between pice changes and positions. The key casual elationship is how past etuns influence cuent tades. In a well functioning maket we should find a negative casual elationship when pices go up, maket paticipants sell and when pices go down, maket paticipants buy see figue 1. If LBHs ae pesent we would expect the opposite casual elationship, a positive feedback tading effect whee pice ises causes moe puchases and pice falls cause moe sales see figue 2. To sepaate these elationships we use a diect Gange-causality analysis using a two vaiable autoegession of maket positions and etuns (see figue 3). This is a well-established pocedue made popula by J. Hasbouck (Jounal of Finance, 1991). Figue 1: No liquidity black holes Endogenous vaiables Retun Signed tade Explanatoy vaiables Past etuns Past signed tades + + Page 6 of 9

7 Figue 2: A maket with positive-feedback tading and liquidity black holes Endogenous vaiables Retun Signed tade Explanatoy vaiables Past etuns? / + Past signed tades + + Figue 3: t = i = 1 α i t i + i = 0 β i x t i + ε 1 t x t = i = 1 γ i t i + i = 1 δ i x t i + ε 2 t t = etun fom t-1 to t xt = 1 if buye initiated tade at t, 0 if no tade at t, -1 if selle initiated tade at t Data To cay out ou empiical tests we use publicly available, weekly data fo five yeas of puchases and sales of futues contacts on the Chicago Boad of Tade, fo 10 majo maket instuments 5 cuencies vesus the dolla, 3 slices of the US bond maket and 2 majo equity makets. This data is moe extensive in the beadth of instuments coveed and length of data seies than used in some elated studies on maket liquidity, but it poses a few hudles. Ideally we would examine daily position data on the spot maket. That said, given the stong coincident and lagged coelation fom daily odes to daily pices, we would be moe likely to find spuious coelation fom pices to odes using daily data. At a weekly level it is highly unlikely that the elationship between odes to data spills ove into the elationship between pices to odes. In this egad, using weekly data is moe obust than using daily data. In a simila vein using futues pices instead of spot pices should also wok against a spuious positive coelation. Futues pices might be expected to exhibit less positive feedback than spot pices because using futues is a lowe Page 7 of 9

8 isk and less capital intensive way of expessing the stabilising view that the cuent maket move is elated to nonfundamental factos and will eventually be unwound. A positive coefficient on a weekly basis would be, theefoe, all the moe supising and moe significant to see. Results Thee is a statistically significant positive co-efficient between past etuns and cuent positions in the case of all of the foeign exchange instuments and a statistically significant negative coefficient fo the bond instuments. The equity instuments show signs of becoming less negative ove time. See table 2. Table 2: The statistical significance of the Gange causality vecto auto-egession of last week s pice to this week s odes T - Statistic 1992/93 - Jan 97 Jan 97 - Jun 99 Jun 99 - Jun 02 CAD CHF JPY AUD MXN NASDAQ y Teasuy note y Teasuy note SP y Teasuy note Aveage FX Aveage Equity Aveage Fixed Income Implications and conclusions The esults in table 2 show supisingly stong evidence that liquidity black holes ae pevalent in cuency makets, not so much, but inceasingly so in equities and not in bond makets. Within the cuency makets the dolla/yen exchange ate, one of the lagest exchange ates in tems of tunove appeas ife with liquidity black holes. The smalle Austalian dolla also exhibits liquidity black holes, but less so than the yen. The esult ae boken up into thee peiods in ode to isolate the impact of the last majo cuency cisis, the Asian financial cisis. Page 8 of 9

9 These esults imply that divesity is key to liquidity and yet it continues to eceive little diect attention by those tying to suppot financial maket liquidity. Indeed the actions of financial maket egulatos - chaged with ensuing we have smooth functioning makets - ae often, indiectly and inadvetently, to educe divesity. It is inteesting that the cuency makets exhibit moe black holes than bond makets. Not least because many of the foces which have educed divesity in the cuency makets ae emeging in bond makets. Thee is inceasing consolidation in industy playes and exchanges. Thee is a woying tend by egulatos to insist that the maket sensitive isk management systems used by the banks big playes in foeign exchange makets ae also used by insuance companies big playes in the bond makets. Divesity mattes and pobably moe so than size in the development of liquid financial makets. Makets can be lage, but pone to toublesome liquidity black holes if they ae not divese. This can be seen in the foeign exchange makets today and thee ae woying signs that divesity is falling in othe majo financial makets. Divesity elates to numbes of playes and instuments, but in makets pone to heding, a citical ole is played by the divesity of decision ules. Maket-sensitive isk management systems educe divesity of decision ules and the encouagement by egulatos fo these systems to be used acoss industy playes will incease the numbe of liquidity black holes. This is avoidable. Good isk management does not mean the use of the same decision ules. Indeed, investos with diffeent investment objectives ae ill seved by isk-management systems which lead them to behave in the same way as othe investos. The manta of egulatos should not be high and common standads, but high and divese standads. In this way egulatos may take advantage of the obsevation that if poo divesity can make lage makets illiquid, then good divesity can make small makets moe liquid than thei size would suggest. Page 9 of 9