Financial Market Infrastructures - Risksubscriptions.risk.net/.../JournalofFinancialMarketInfrastructures.pdf · The Journal of Financial Market Infrastructures EDITORIAL BOARD Editor-in-Chief

Volume 5 Number 1September 2016

The Jo

urn

al of Fin

ancial M

arket Infrastru

ctures

Volume 5 N

umber 1 Septem

ber 2016

■ The role of collateral in supporting liquidityYuliya Baranova, Zijun Liu and Joseph Noss

■ The effective supply of collateral in AustraliaBelinda Cheung, Mark Manning and Angus Moore

■ Mobilization of collateral in Germany as a reflection of monetary policy and financial market developmentsAlexander Müller, Jan Paulick, Jan Fichtner and Hubert Wittenmayer

■ Collateral flows and balance sheet(s) spaceManmohan Singh

■ Impact of monetary policy on collateral reuseAmeya Muley

■ Collateral chains and incentivesCharles M. Kahn and Hyejin Park

The Journal of

Special IssueCollateralGuest Editor: Manmohan Singh

Financial MarketInfrastructuresPEFC Certified

This book has been produced entirely from sustainable papers that are accredited as PEFC compliant.

www.pefc.org

JFMI_5_1_Spetember_2016.indd 1 26/08/2016 16:13

Tria

l Cop

y For all subscription queries, please call:

UK/Europe: +44 (0) 207 316 9300

USA: +1 646 736 1850 ROW: +852 3411 4828

in numbers

140,000

Users

Page views

19,400+ on Regulation

6,900+ on Commodities

19,600+ on Risk Management

6,500+ on Asset Management

58,000+ articles stretching back 20 years

200+

New articles & technical papers

370,000

21,000+ on Derivatives £

Visit the world’s leading source of exclusive in-depth news & analysis on risk management, derivatives and complex fi nance now.

(each month)

(each month)

See what you’re missing

(each month)

RNET16-AD156x234-numbers.indd 1 21/03/2016 09:44

To subscribe to a Risk Journal visit Risk.net/subscribe or email [email protected]

http://subscriptions.risk.net/subscribe/

mailto:corpsubs%40risk.net?subject=

The Journal of Financial Market InfrastructuresEDITORIAL BOARD

Editor-in-ChiefRon Berndsen De Nederlandsche Bank and Tilburg University

Associate EditorsSujit Chakravorti The Clearing HouseMassimo Cirasino World BankRodney Garratt University of CaliforniaTerry Goh Monetary Authority of SingaporeGerard Hartsink Global Legal Entity

Identifier FoundationRichard Heckinger Federal Reserve Bank

of Chicago (former)Ronald Heijmans DNBLex Hoogduin University of AmsterdamCharles Kahn University of IllinoisThorsten Koeppl Queen’s UniversityEsmond Lee Hong Kong Monetary AuthorityGottfried Leibbrandt SWIFTCarlos León Central Bank of ColombiaKlaus Löber European Central Bank

Mark Manning Reserve Bank of AustraliaAlistair Milne Loughborough UniversityMasayuki Mizuno Bank of JapanDaniela Russo European Central BankEdwin Schooling Latter Financial

Conduct Authority (UK)Manmohan SinghJeff Stehm Promontory Financial GroupLawrence Sweet Federal Reserve Bank of

New YorkJohn Trundle Euroclear UK & IrelandLeo Van Hove Free University of BrusselsWolf Wagner Erasmus UniversityStuart E. Weiner Santa Fe Group and

Dartmouth CollegeFroukelien Wendt

SUBSCRIPTIONSThe Journal of Financial Market Infrastructures (Print ISSN 2049-5404 j Online ISSN 2049-5412)is published quarterly by Incisive Risk Information Limited, Haymarket House, 28–29 Haymarket,London SW1Y 4RX, UK. Subscriptions are available on an annual basis, and the rates are set outin the table below.

UK Europe USRisk.net Journals £1945 €2795 $3095Print only £735 €1035 $1215Risk.net Premium £2750 €3995 $4400

Academic discounts are available. Please enquire by using one of the contact methods below.

All prices include postage.All subscription orders, single/back issues orders, and changes of addressshould be sent to:

UK & Europe Office: Incisive Media (c/o CDS Global), Tower House, Sovereign Park,Market Harborough, Leicestershire, LE16 9EF, UK. Tel: 0870 787 6822 (UK),+44 (0)1858 438 421 (ROW); fax: +44 (0)1858 434958

US & Canada Office: Incisive Media, 55 Broad Street, 22nd Floor, New York, NY 10004, USA.Tel: +1 646 736 1888; fax: +1 646 390 6612

Asia & Pacific Office: Incisive Media, 20th Floor, Admiralty Centre, Tower 2,18 Harcourt Road, Admiralty, Hong Kong. Tel: +852 3411 4888; fax: +852 3411 4811

Website: www.risk.net/journal E-mail: [email protected]




The Journal of Financial Market InfrastructuresGENERAL SUBMISSION GUIDELINES

Manuscripts and research papers submitted for consideration must be original workthat is not simultaneously under review for publication in another journal or otherpublication outlets. All articles submitted for consideration should follow strict aca-demic standards in both theoretical content and empirical results. Articles should beof interest to a broad audience of sophisticated practitioners and academics.

Submitted papers should follow Webster’s New Collegiate Dictionary for spelling,and The Chicago Manual of Style for punctuation and other points of style, apart froma few minor exceptions that can be found at www.risk.net/journal. Papers should besubmitted electronically via email to: [email protected]. Please clearlyindicate which journal you are submitting to.

You must submit two versions of your paper; a single LATEX version and a PDFfile. LATEX files need to have an explicitly coded bibliography included. All files mustbe clearly named and saved by author name and date of submission. All figures andtables must be included in the main PDF document and also submitted as separateeditable files and be clearly numbered.

All papers should include a title page as a separate document, and the full names,affiliations and email addresses of all authors should be included. A concise andfactual abstract of between 150 and 200 words is required and it should be includedin the main document. Five or six keywords should be included after the abstract.Submitted papers must also include an Acknowledgements section and a Declarationof Interest section. Authors should declare any funding for the article or conflicts ofinterest. Citations in the text must be written as (John 1999; Paul 2003; Peter and Paul2000) or (John et al 1993; Peter 2000).

The number of figures and tables included in a paper should be kept to a minimum.Figures and tables must be included in the main PDF document and also submittedas separate individual editable files. Figures will appear in color online, but willbe printed in black and white. Footnotes should be used sparingly. If footnotes arerequired then these should be included at the end of the page and should be no morethan two sentences. Appendixes will be published online as supplementary material.

Before submitting a paper, authors should consult the full author guidelines at:

http://www.risk.net/static/risk-journals-submission-guidelines

Queries may also be sent to:The Journal of Financial Market Infrastructures, Incisive Media,Haymarket House, 28–29 Haymarket, London SW1Y 4RX, UKTel: +44 (0)20 7004 7531; Fax: +44 (0)20 7484 9758E-mail: [email protected]




The Journal of

Financial MarketInfrastructures

The journalToday, in the light of the financial crisis, it has become part of the political agendato strengthen payment, clearing and settlement systems, as well as repositories fordata on the trades they process. In 2012 a new set of internationally agreed CPSS-IOSCO Principles carved out financial market infrastructures (FMIs) as a distinctarea in financial policy. The Journal of Financial Market Infrastructures is the firstjournal to focus on this exciting and dynamic sector, and aims to bring together acommunity of contributors from the constituent sectors to analyze FMIs to further thedevelopment of this emerging field. The journal provides a balanced representation ofacademic and practitioner-focused papers that are dedicated to analyzing operationaland regulatory effectiveness and efficiency of payment, clearing, settlement and traderepository systems; and the risks they manage, transmit and create.

The Journal of Financial Market Infrastructures considers submissions in theform of technical papers and policy-oriented papers (forum discussions) on topicsincluding, but not limited to, the following:

� systemically important payment systems,� securities settlement systems,� central counterparties,� central securities depositories,� trade repositories,� settlement risk and other FMI-related risks including interdependencies,� infrastructure-related systemic risk,� network analysis of an FMI,� critical service providers and non-bank payment service providers,� correspondent banking,� FMI liquidity and collateral management,� exchanges and multilateral trading platforms,� oversight and supervision of FMIs, and� FMI-related standardization and legislation.




The Journal of Financial Market Infrastructures Volume 5/Number 1

Special Issue: Collateral

CONTENTS

Letter from the Guest Editor vii

The role of collateral in supporting liquidity 1Yuliya Baranova, Zijun Liu and Joseph Noss

The effective supply of collateral in Australia 27Belinda Cheung, Mark Manning and Angus Moore

Mobilization of collateral in Germany as a reflection of monetary policyand financial market developments 49Alexander Müller, Jan Paulick, Jan Fichtner and Hubert Wittenmayer

Collateral flows and balance sheet(s) space 65Manmohan Singh

Impact of monetary policy on collateral reuse 83Ameya Muley

Collateral chains and incentives 103Charles M. Kahn and Hyejin Park

Editor-in-Chief: Ron Berndsen Subscription Sales Manager: Aaraa JavedPublisher: Nick Carver Global Key Account Sales Director: Michelle GodwinJournals Manager: Dawn Hunter Composition and copyediting: T&T Productions LtdEditorial Assistant: Carolyn Moclair Printed in UK by Printondemand-Worldwide

©Copyright Incisive Risk Information (IP) Limited, 2016. All rights reserved. No parts of this publicationmay be reproduced, stored in or introduced into any retrieval system, or transmitted, in any form or by anymeans, electronic, mechanical, photocopying, recording or otherwise without the prior written permission of thecopyright owners.




SPECIAL ISSUE: COLLATERAL

LETTER FROM THE GUEST EDITORManmohan Singh

This special issue of The Journal of Financial Market Infrastructures on collateralprovides a collection of papers that are not fully in sync with traditional thinking inpolicy and academic circles (such as collateral reuse rate), or are under-researched. Myhope is that this special issue will provide some food for thought for policy makersand academics as the financial markets and central banks, as well as forthcomingregulations, continue to change the demand for and supply of available collateral.

The issue’s first paper, by Yuliya Baranova, Zijun Liu and Joseph Noss from theBank of England, finds that a reduction in the willingness of market participants toact as intermediaries in collateral markets will likely lead to adverse consequencesfor market functioning and financial stability.

Our second paper is from the Reserve Bank of Australia and is one of the first toestimate the collateral reuse rate in Australia via a survey on collateral activity. Theauthors, Belinda Cheung, Mark Manning and Angus Moore, find that the reuse rateis lower now, in 2016, than it was in 2014.

The third paper in the issue, from Alexander Müller, Jan Paulick, Jan Fichtner andHubert Wittenmayer from Deutsche Bundesbank, provides a snapshot of collateralsubmitted from 2008 to 2016 via different mobilization channels, and summarizeshow submission of collateral shifted during times of crisis and then again due tomonetary policy decisions.

The issue’s fourth paper, from Manmohan Singh, reiterates that collateral does notflow through a vacuum; it needs (on- or off-) balance sheet space to move within thefinancial system, and ad hoc allocation of balance sheet space by central banks is notconducive to monetary policy transmission.

In the past few years, in several dissertations (including job market papers) andpolicy papers, researchers have tried to provide macro-foundations for the financialplumbing of the market. An example is our fifth paper by Ameya Muley from the MITeconomics department (it forms part of his dissertation) and it provides the theorybehind the plumbing, concluding that when central banks remove good collateral fromthe market this leads to lower investment and lower aggregate output in the economy.Thus, from a policy perspective, exhausting policy rates first (including below-zerorates) may be more effective than siloing good collateral when scarce.

The issue’s final paper is from Charles M. Kahn and Hyejin Park from the Uni-versity of Illinois at Urbana-Champaign. The authors model the asymmetry betweencollateral values to the parties in the collateral chain: a reason why hedge funds resortto rehypothecation of their assets with their prime broker rather than an outright sale

www.risk.net/journal




in the market to fund their positions. In this more analytical paper it is highlightedthat collateral reuse can be socially beneficial if the costs of misallocation are notsignificant.

I hope the readers of The Journal of Financial Market Infrastructures will attempt tojourney into uncharted waters, especially since global financial markets are currentlytruly at a crossroads.

Journal of Financial Market Infrastructures 5(1)




Journal of Financial Market Infrastructures 5(1), 1–26DOI: 10.21314/JFMI.2016.063

Research Paper

The role of collateral in supporting liquidity

Yuliya Baranova, Zijun Liu and Joseph Noss

Bank of England, Threadneedle Street, London EC2R 8AH, UK;emails: [email protected]; [email protected];[email protected]

(Received May 16, 2016; revised June 10, 2016; accepted June 10, 2016)

ABSTRACT

Collateral plays an important role in supporting a vast range of transactions thathelp ensure the efficient functioning of the financial system. But collateral marketsalso have the potential to exacerbate risks to financial stability, not least given that,during periods of market stress, demand for high-quality collateral may increase,while collateral availability may fall. This paper offers a means to estimate howthis potential imbalance between collateral supply and demand is likely to vary as afunction of market stress. In addition, it offers an estimate of the increase in marketvolatility sufficient to cause a dislocation in the market for collateral and a subsequentdeterioration in market functioning. It suggests that, from the perspective of financialstability, the implications of an imbalance between the supply and demand of collateralare likely to be comparatively benign, but that a reduction in the willingness and/orability of market participants to act as intermediaries in collateral markets is likely tohave more serious consequences for market functioning. This work also provides aframework through which policy makers can investigate how regulations might affectthe proximity of these risks.

Keywords: collateral; securities-financing transactions; derivatives; regulation; liquidity.

Corresponding author: J. Noss Print ISSN 2049-5404 j Online ISSN 2049-5412Copyright © 2016 Incisive Risk Information (IP) Limited

1




2 Y. Baranova et al

1 INTRODUCTION

Collateral, ie, securities pledged to secure loans and other counterparty exposures, isused to support a vast range of market-based transactions that together help supportthe efficient functioning of the financial system. It plays a key role in mitigatingcounterparty credit risk, and acts as a store of liquidity with which firms can managetheir funding. It also plays an important role in enabling leveraged financial institutionsto fund purchases (and collateralize against short sales) and financial assets. This latterrole is particularly important in supporting market liquidity, which in turn helps toensure the provision of market-based finance (Anderson et al 2015). In what follows,we refer to the role of collateral in supporting such activities as that of supporting“liquidity”.

The behavior of collateral markets can also exacerbate risks to financial stability.One such risk arises from the possible procyclical behavior of the balance betweenthe demand for collateral and the market participants’ ability and/or willingness tosupply it. During the recent financial crisis, the quantity of high-quality securitiesmade available for use as collateral reduced, due to perceptions of increased coun-terparty credit risk (which caused a reduction in securities lending) and as marketintermediaries sought to deleverage. At the same time, demand for high-quality col-lateral increased, as, for example, firms began to hoard liquid assets, volatile marketscaused an increase in margin held against derivatives and lower-quality securitieswere no longer accepted as collateral.1 Together, these developments contributed to apernicious spiral of rising margin requirements, lower risky asset prices and decliningmarket functioning and liquidity (including in markets that played an important rolein extending funding to the real economy).

Given these developments, it is perhaps not surprising that recent regulation hassought to reduce some of the procyclicalities associated with collateral. Such regu-lation includes a set of numerical haircut floors for noncentrally cleared securities-financing transactions, and the agreement of methodological standards around howthey should be set (see Financial Stability Board 2014). The regulatory minimumleverage ratio (see Basel Committee on Banking Supervision 2014) also has the poten-tial to limit the extent to which financial intermediaries can increase their capacityduring periods of benign market conditions. And, in derivatives markets, most juris-dictions are in the process of adopting margin requirements for noncentrally clearedderivatives (see Financial Stability Board 2015). Together, these regulations shouldgo some way toward dampening the procyclicalities seen during the crisis, and limit

1 For a more in-depth account of these developments, see Berrospide (2012) and Gorton and Metrick(2012).

Journal of Financial Market Infrastructures www.risk.net/journal




The role of collateral in supporting liquidity 3

the degree to which the effective supply of collateral expands during benign marketconditions, only to contract sharply during stress.

There remains, however, the important question of the level(s) at which the effectivesupply of, and demand for, high-quality collateral will settle following the introductionof these regulations, and whether this will be sufficient to ensure the stable functioningof financial markets during periods of stress.2 This is particularly important, given thatsome of the regulation introduced since the crisis, while reducing the procyclicalitiesdescribed above, might also increase demand for high-quality collateral during periodsof stress (see Financial Stability Board 2014).

To assess this question, we begin by providing a comprehensive framework toassemble and quantify the factors that affect the supply of, and demand for, high-quality collateral.Although the aggregate supply of such collateral is vast, only a smallproportion of this is made available – via securities-lending and repo transactions –to support market functioning. We then estimate the degree to which this availablesupply of high-quality collateral is likely to decrease, and its demand increase, duringperiods of stress. This decrease occurs for a number of reasons, including the deteri-oration in market participants’ perceptions of their counterparties’ creditworthiness,and increases in the demand for collateral for use as initial margin and in liquid assetbuffers. We capture this variation in supply and demand by drawing together a numberof empirical relationships between market participants’behavior in collateral marketsand a combination of average dealer credit default swap premiums (which we takeas a proxy for perceptions of counterparty credit risk) and the VIX volatility index(which we take as a proxy for market stress and general economic conditions).3

We also consider the role of intermediaries in the market for high-quality collateral.There is some evidence that during the recent crisis such intermediation activityby broker-dealers reduced considerably, as falls in asset prices forced such firmsto deleverage. Given the lack of historical data on such dynamics, we provide atheoretical model of this behavior – parameterized using broker-dealer balance sheetdata – and use this to estimate the degree to which the intermediation in collateralmarkets might be reduced during periods of stress.

This modeling has two key results. First, there is a risk that demand for high-qualitycollateral exceeds the supply made available during periods of stress. Our analysisindicates that this risk might crystallize at levels of stress commensurate with theVIX volatility index exceeding around 44% for a period of around a quarter. This hasoccurred on only one occasion historically, in late 2008, at the height of the recent

2 In what follows, we implicitly assume this equilibrium level to be nonzero, ie, that a cash-onlysystem would be inefficient, either because some leverage is a requirement of an efficient financialsystem, or because it is inefficient for institutions that are not cash rich to hold cash as an asset.3 This is not without precedent (see Rey 2013).

www.risk.net/journal Journal of Financial Market Infrastructures




4 Y. Baranova et al

crisis. That said, even if this risk were to crystallize, we judge that it might have onlya moderate impact, as the cost of borrowing high-quality collateral might, in time,adjust to restore equilibrium between supply and demand.

A second risk – whose impact we judge to be potentially more severe – arises frommost end-users of collateral not being directly connected to its suppliers. Instead, anetwork of leveraged financial institutions (principally broker-dealers) acts as inter-mediaries between those that supply high-quality collateral and those that demand it.As market stress intensifies, the ability and willingness of these institutions to obtainthe leverage necessary to perform this intermediation tends to decrease. This leadsto the risk that, in future periods of stress, although the demand for collateral mightnot exceed unleveraged end-investors’ ability to supply it, collateral may become“blocked”: that is, unable to reach those that wish to use it, due to a shortage ofintermediation capacity.

While we estimate this second risk might crystallize at similar levels of marketstress to the first risk, its consequences for financial stability could be much morefar-reaching. These might include market participants’ sudden inability to obtain thecollateral they need to manage the risks associated with their business, includingpayment of initial margin on derivatives transactions, and impairing dealers’ abilityto fund other leveraged investors’ (ie, hedge funds’) purchases of financial assets,which might have implications for market functioning and liquidity.

Throughout, our analysis is informed by a range of background literature. Thisprovides partial analysis of many of the dynamics that have increased the relevance ofcollateral to financial stability in recent years. Krishnamurthy and Vissing-Jorgensen(2012) examine the increase in demand for high-quality collateral, particularly UStreasuries, during the recent crisis. Pozsar (2011) asks whether the supply of suchsecurities might be insufficient to meet institutional investors’ demand for safe andliquid instruments.Aguiar et al (2016) provides a map of collateral uses and flows, anddocuments an increased demand for collateral due to regulatory reforms. Singh (2011)examines the frequency with which collateral is reused and documents a decline inboth the supply of collateral by end users and the “velocity” with which it passedbetween intermediaries once the financial crisis had peaked. A number of papersalso examine collateral reuse in a number of jurisdictions (see, for example, Fuhreret al 2015; Cheung et al 2014; Capel and Levels 2014). Other recent literature (eg,European Systemic Risk Board 2014) has focused on how the increased reliance oncollateral may pose financial stability risks. Infante (2015) uses a theoretical model toshow that dealers acting as intermediaries in the market for collateral can be exposedto the risk of a sharp withdrawal of funding by collateral providers.

This paper extends the existing literature by drawing together many of these dynam-ics, to provide a comprehensive framework with which to assess the drivers of the






supply of, and demand for, collateral during periods of stress. Its applications to publicpolicy are broadly threefold.

(1) Our results provide a ready reckoner for the degree of market stress that might– according to the model offered here – be expected to trigger an imbalancebetween the supply of, demand for and insufficient intermediation of collateral.Our framework may therefore serve as a quantitative risk assessment tool forany authority interested in assessing risks associated with collateral markets.

(2) Our framework can be used to consider how changes to the structure of thefinancial system – and, in particular, the network of market intermediaries whostand between the largely unleveraged institutions acting as end suppliers andusers of collateral – might change the level of market stress at which these risksare likely to crystallize.

(3) Policy makers could use this framework to consider how recently introducedregulations (including the leverage ratio, initial margin and minimum haircutrequirements) might affect market participant behavior, and hence the level ofmarket stress at which these risks may crystallize.

Throughout, the focus of our analysis is confined to high-quality collateral.4 Thisis because, during past periods of market stress, market participants have tended toturn to higher-quality securities for use as collateral. In particular, we assume thatonly high-quality securities are used as initial margin, as central counterparty defaultfund contributions and in bank liquid asset buffers.5

This paper proceeds as follows. Section 2 offers a framework for considering thekey drivers of the supply of, and demand for, high-quality collateral, and a guide totheir current levels (at least following the full implementation of recently introducedregulation). Section 3 considers how such supply and demand may vary with marketstress, eg, as a result of the prudent risk management of market participants, and therisks to which these give rise. The role of leveraged intermediaries, including broker-dealers, is considered in Section 4, together with the further risks that arise from theiractivities. Section 5 concludes. Technical details are confined to the annex, availableonline.

4 This includes AAA/AA rated government bonds (excluding China), agency mortgage-backedsecurities (MBSs) and supranational institutions.5 The definition of high-quality assets applied here is more narrow than that accepted in banks’liquidity coverage ratio under recently agreed international regulation (see Basel Committee onBanking Supervision 2013).





6 Y. Baranova et al

TABLE 1 Aggregate supply of high-quality collateral.

Amount outstandingType of security (US$ trillion)

AAA/AA rated government bonds (excluding China) 34.6Agency mortgage backed securities 6.0Supranationals 1.3

Total 41.8

Source: Bank for International Settlements (BIS), Securities Industry and Financial Markets Association (SIFMA)and Dealogic. Data as of September 2014. Values in the table do not add up due to rounding errors.

2 A STYLIZED FRAMEWORK FOR CONSIDERING THESTEADY-STATE SUPPLY/DEMAND OF COLLATERAL

This section sets out a framework for considering the components of collateral sup-ply and demand. We assume that the demand for collateral is exogenously drivenby regulatory requirements and demand for safe assets, as described in more detailbelow. In the absence of supply constraints, prices should adjust so that the steady-state supply always equals demand. In practice, however, the supply of collateral isactually bounded above by the amount of collateral made available to borrowers (the“made available supply”), which is not currently binding according to our estimates,but could become so as market stress increases, as described in Section 3. In thefollowing, we estimate the made available supply of collateral in a number of steps,starting from the aggregate supply and unencumbered supply of collateral.

The analysis is predicated on benign market conditions, which we take to corre-spond to those in the three months ending in August 2015, when the VIX volatilityindex averaged around 17%. The values in Table 1 apply to the global market forhigh-quality collateral, and represent estimates of the supply of, and demand for,collateral following the full implementation of recently introduced regulation.

2.1 Aggregate supply of collateral versus unencumbered supply ofhigh-quality collateral

The aggregate supply of high-quality collateral (ie, the amount of high-quality secu-rities outstanding) is vast, and at end-2014 was estimated to be in the region ofUS$42 trillion (Table 1).

Only a small proportion of this is available to support market functioning, however.This is because a large proportion of the total supply of high-quality collateral isencumbered, ie, it is in some sense siloed and used for a purpose that prevents it






TABLE 2 Encumbrance of high-quality collateral (in trillions of US dollars).

Amount Source of UnencumberedOwner type Holdings encumbered encumbrance supply

Governmentalinstitution

8.9 8.9 Inability toengage insecuritieslending

0.0

Commercialbank

5.3 4.5 Liquid assetbuffer or initialmargin

0.8

Insurancecompany orpension fund

5.7 0.0 5.7

Centralbanks�

4.4 4.2 Mostlylendingagainst othergovernmentbonds

0.2

Non-resident��

11.5 11.3 Foreignexchangereserves

0.2

Other�� 6.0 3.5 Various 2.6

Total 41.8 32.3 9.5

Total postderivativesreform

33.3 8.5

Source: BIS, SIFMA, European Central Bank (ECB), International Monetary Fund (IMF). Numbers may not add updue to rounding. �Includes securities purchased through quantitative easing programs. ��Refers to owners of high-quality collateral domiciled in jurisdictions other than those in which high-quality collateral is originated. ��Includesmoney market funds, asset managers and other non-bank financial institutions.

being used to support liquidity.6 This occurs for a variety of reasons, including use ininstitutions’ liquid asset buffers and initial margins/default fund contributions againstderivative exposures, as well as siloing by end-investors, who are prevented fromlending securities to other investors.7 These types of collateral encumberance we treatas fixed claims that, for low levels of market volatility, reduce the available collateral

6 This definition of “encumbrance” differs to that used in recent financial risk assessment andregulation (see Bank of England 2012), in which an asset is said to be “encumbered” if it is underclaim by another party.7 For example, some intragovernmental holdings, government foreign exchange reserves, centralbank and money market fund assets are unavailable for use in securities lending.





8 Y. Baranova et al

TABLE 3 Split of total stock of high-quality securities made available for loan, and on loan,by type of beneficial owner.

Securities available SecuritiesType of owner for loan (%) on loan (%)

Pension fund and insurers 39 34Banks 11 15Sovereign wealth funds 24 32Asset managers 20 12Other financial 4 6Nonfinancial 2 1

Total 100 100

Source: Markit.

pool (the possibility that they increase as a result of market stress is examined inSection 3).

Estimates of the total stock of high-quality securities encumbered for the reasonsabove are given in Table 2. The remaining unencumbered supply of high-quality col-lateral lies at US$9.5 trillion. This figure does not, however, account for the additionalencumbrance of collateral due the implementation of recently introduced over-the-counter derivatives regulatory reforms, which include the mandatory exchange ofmargin on a variety of derivatives transactions. According to Bank for InternationalSettlements (2013), the full implementation of such reforms will result in roughlyUS$1 trillion of additional high-quality collateral being encumbered. This has theeffect of reducing the unencumbered supply to US$8.5 trillion: US$7.7 trillion heldby largely unleveraged institutions (who engage in securities lending (see below))and US$0.8 trillion held by dealers.

2.2 Unencumbered supply versus made available supply ofhigh-quality collateral

Only a fraction of the unencumbered supply of collateral is made available for use infacilitating market transactions, as explained below. In what follows we refer to thisas the “made available supply” of collateral. This comes about via two main channels.

(1) Securities lending: in which beneficial owners make their assets available fordealers to borrow. These institutions are typically long-term investors thatuse securities lending to generate additional income on their portfolios. Theyinclude insurance companies, pension funds, asset managers, sovereign wealthfunds, commercial banks and other financial and nonfinancial institutions.






Table 3 gives a breakdown of high-quality securities that are available for loanand those that are on loan, by type of beneficial owner.

(2) Dealer financing: whereby banks and dealers who hold high-quality securitiesrepo them out to finance their operations. We estimate collateral involved insuch activity to stand at around US$773bn.8;9

We estimate that, at present, only 28% of high-quality assets held by owners partici-pating in securities-lending programs (the largest constituent of total unencumberedsupply) are actually available for loan. This is likely due in part to the regulatoryconstraints on institutional investors that prevent them from engaging in securities-lending transactions on a larger scale.10 Such securities lending is also likely to varyover the economic cycle, eg, as the result of beneficial owners’ changing perceptionsof counterparty default risk (a dynamic explored further in Section 3).

2.3 Demand for high-quality collateral

Not all of the made available supply of high-quality collateral is utilized, however. Weestimate that the total demand for high-quality collateral currently stands at aroundUS$1.6 trillion.11 This comes from a number of sources, including reverse repos (ie,transactions in which cash is lent against high-quality collateral) performed by realmoney investors (including money market funds), and lending of lower-quality secu-rities against high-quality collateral (both of which are assumed to remain constantover time).12;13 Demand for high-quality collateral also stems from the reinvestment of

8 Collateral supply provided via dealer financing is calculated as the difference between currentend-user demand for high-quality collateral (as given in Table 4) and the amount of high-qualitycollateral on loan.9 There are two other channels through which the unencumbered supply could be made availablefor loan: (i) fund leverage, whereby investment funds with long positions in high-quality collateralmay also choose to gain leverage by pledging the collateral with dealers in the repo markets andthereby securing financing for other asset purchases; and (ii) assets placed with dealers by theirclients (eg, hedge funds) for a variety of purposes. Both activities are, in the case of high-qualitycollateral, relatively small in scale, and so are ignored in what follows.10 For example, according to Financial Stability Board (2012), in the United States, registeredinvestment companies (which include mutual funds, money market mutual funds, closed-end fundsand exchange-traded funds) may not lend more than one-third of their total assets under management.11 Note that high-quality collateral used as initial margins and banks’ liquid asset buffers in benignmarket conditions is, under this framework, considered part of the encumbered collateral supply.12 While the magnitude of money market funds’ reverse repos might be expected to increase duringmarket stress, we see little evidence of this empirically. And given that those institutions thatreceive high-quality collateral against the lending of lower-quality securities are exposed to littlecounterparty risk in doing so, we might expect this activity to stay constant across the cycle.13 Collateral siloed within payment and settlement systems is not included in our framework due toits limited size and procyclicality.





10 Y. Baranova et al

TABLE 4 Sources of demand for high-quality collateral.

Current demandSource (US$ bn)

(i) Reverse repos by real money investors 1261(ii) Lending of non-HQ securities against HQ collateral 184(iii) Reinvestment of cash collateral by securities lenders 176

Total 1621

Source: Data Explorers.

FIGURE 1 A summary of the components of the global supply of, and demand for,high-quality collateral.

0123456789

05

1015202530354045

US

$ (t

rillio

n)

US

$ (t

rillio

n)

34.6

6.01.3Total

supply(LHS)

Madeavailablesupply(RHS)

Demand(RHS)

Encumbered(LHS)

Unencumbered(LHS)

Unencumbered(RHS)

Govbonds

Supply Demand

8.9

5.54.2

11.3

3.5

Intragovernmental

Domestic banks(IM/LAB)Central banks

Nonresident(FX reserves)

Other

8.5 Zoom in

Zoom

in

7.7

0.8

2.2

0.8 0.80.8

Beneficial ownersDealers

Source: BIS, SIFMA, Data Explorers, Dealogic, ECB and IMF. Numbers may not add up due to rounding.

cash collateral by securities lenders (in particular, in reverse repos against high-qualitycollateral). Table 4 provides a summary.

We assume that collateral demanded at this stage is not reused further. In particular,money market funds and corporates investing cash in reverse repo do not reuse thehigh-quality collateral that they receive. Also, agent lenders (largely banks) that rein-vest cash in reverse repo against high-quality assets on behalf of beneficial ownersdo not have the legal right to reuse this collateral.

2.4 Summary of analytical framework

A summary of the components of the supply of, and demand for, high-quality collateralis shown in Figure 1. Note that, of an overall US$41.8 trillion supply of high-quality






collateral (far left bar), we estimate that only US$1.6 trillion is used on a day-to-daybasis to support market functioning (far right bar).

3 HOW THE SUPPLY/DEMAND OF HIGH-QUALITY COLLATERALMIGHT VARY WITH MARKET STRESS

This section provides a simple and stylized model of how changes in the supply of,and demand for, high-quality collateral vary with market stress. In doing so, it aimsto assess the risk that the demand for such collateral might exceed its supply duringperiods of financial market turbulence.

The overall approach is based on a simple model of market participant behavior,and how this varies as a function of two variables that together are used to capturethe degree of stress in financial markets:

� the VIX volatility index;14

� the average credit default swap (CDS) premiums on the senior unsecured debtof global systemically important banks (GSIBs),15 which we take as a proxyfor perceptions of counterparty risk associated with major dealers.

For presentational convenience, we show our results as a function of a single variable,the VIX index. To do so, it is necessary to estimate a mapping between this andthe average CDS premiums. This historical relationship between the VIX index andCDS premiums varies considerably over time and may have been altered by theeffect of recently introduced regulation, which aimed to increase the resilience (andhence perceptions of counterparty risk) of dealers during stress. Therefore, ratherthan drawing on the historical observed relationship between market volatility andfinancial institutions’CDS premiums (which might not be representative of that likelyto be held in future), we employ a structural model of credit risk and show how thisvaries with the volatility of dealers’ assets.

Throughout, we aim to capture behavior that is consistent with market participants’prudent risk management, ie, the degree to which they scale back their securitieslending, demand more initial margin and increase their liquid asset buffers, as anincreasing function of market stress and counterparty credit risk. Actual behaviors

14 TheVIX index is a measure of market expectations of thirty-day volatility, as conveyed by Standard& Poor’s 500 (S&P 500) stock index options prices. It is widely taken as a summary measure formarket participants’ perceptions of uncertainty and risk aversion (see, for example, Rey 2013).15 GSIBs included are HSBC, JP Morgan, Barclays, BNP Paribas, Citigroup, Deutsche Bank, BankofAmerica, Credit Suisse, Goldman Sachs, Mitsubishi, Morgan Stanley and Royal Bank of Scotland.Other GSIBs are not included due to their limited presence in collateral markets.






FIGURE 2 Level of high-quality securities available for loan and average dealer CDSpremiums.

1.0

1.5

2.0

2.5

3.0

0

50

100

150

200

250

300

2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

Dea

ler

CD

S p

rem

ium

s (b

ps)

Hig

h-qu

ality

sec

uriti

es a

vaila

ble

for

loan

(U

S$

trill

ion)

Dealer CDS premiumsHigh-quality securities available for loan

Source: Bloomberg and Data Explorers.

of investors are likely more complicated and can be more or less procyclical thanthose shown here. Nonetheless, we set out a simple framework illustrating the risksinvolved.

3.1 A simple model of how the made available supply of collateralvaries with perceptions of intermediaries’ counterparty risk

Historically, the supply of collateral made available via securities lending has var-ied (particularly during the crisis) with lenders’ varying perceptions of counterpartycredit risk (see Dive et al 2011). The relationship between high-quality securitiesavailable for loan and counterparty credit risk is shown in Figures 2 and 3. Duringthe crisis, the amount of high-quality securities available for loan clearly fell withincreased perceptions of counterparty risk. While, over the entirety of the sample,the relationship between the two series lacks statistical significance, we nonethelessinclude it here in order to capture the possible directionality of the relationship duringperiods of future stress.






FIGURE 3 Change in average dealer CDS premiums versus change in proportion ofhigh-quality securities available for loan (as a proportion of beneficial owners’ total supplyof unencumbered collateral).

12

10

8

6

4

2

0

–2

–4

–6

–8

Cha

nge

in p

ropo

rtio

n of

sec

uriti

esav

aila

ble

for

loan

(%

)

–100 –50 0 50 100Change in average dealer CDS premiums

2008 Q4

2011 Q42008 Q3

y = –0.0002x + 0.0014

Source: BIS, SIFMA, ECB and Data Explorers.

We capture this co-movement between extreme changes in CDS premiums andsecurities-lending availability (expressed as a proportion of beneficial owners’ hold-ings of high-quality collateral) by means of a simple linear regression, the slope ofwhich is shown by the black line in Figure 3. This suggests that a 10 basis pointincrease in average dealer CDS premiums is associated with a 0.2 percentage pointreduction in the proportion of beneficial owners’ securities available for loan.

3.2 A simple model of how collateral demand varies with marketstress and perceptions of intermediaries’ counterparty risk

There are two sources of variation in demand for collateral, both of which vary withthe level of market stress: reinvestment of cash collateral from securities-lendingtransactions, and changes in initial margins/liquid asset buffers from their levels inbenign market conditions. These are dealt with in turn.

Throughout, we abstract from the possibility that during market stress the value ofhigh-quality collateral is likely to increase (for example, due to increased governmentbond issuance). This may lead to a slight underestimation of the available collateralsupply during stress.






FIGURE 4 Percentage change in amount of securities lent against cash versus changein average dealer CDS premiums.

20

15

10

5

0

–5

–10

–15

–20

Cha

nge

in a

mou

nt o

f sec

uriti

esle

nt a

gain

st c

ash

(%)

–100 100–50 0 50Change in dealer CDS premiums

2008 Q4

y = –0.0942x – 0.6259

Source: Bloomberg and Data Explorers.

3.2.1 Reinvestment of cash collateral in reverse repos againsthigh-quality assets

We assume that the size of cash collateral reinvestment via reverse repo falls duringstress, as lenders become more concerned about counterparty risks. This fall has twodrivers:

� a fall in the amount of securities lent against cash (due to increased counterpartyrisk, proxied by dealer CDS premiums, driving an overall decline in securities-lending activity);

� fall in the proportion of cash collateral reinvested in reverse repos against high-quality securitiesClosing parenthesis added here for balance – OK?, proxiedby the VIX index.

Again, both effects are captured using linear regressions. Figure 4 shows percentagechanges in the amount of securities lent against cash versus the change in dealerCDS premiums, while Figure 5 shows changes in the proportion of cash collateralreinvested in reverse repos against high-quality collateral versus changes in the VIXindex. Again, although the latter relationship lacks statistical significance, we include






FIGURE 5 Change in the percentage of cash collateral reinvested in high-quality reposversus change in the VIX index.

6

4

2

0

–2

–4

–6

–8

Cha

nge

in p

erce

ntag

e of

cas

h co

llate

ral

rein

vest

ed in

rev

erse

rep

os

40–20 –10 0 302010Change in VIX index

2008 Q4

y = –0.1184x – 0.0266

Source: Bloomberg and RMA.

it in order to capture the possible directionality of relative changes in the two variablesduring future periods of stress.

3.2.2 Demand for use of collateral in initial margin and liquid assetbuffers

Demand for collateral for use as initial margin and as part of banks’ liquid assetsis assumed to increase with market stress. In what follows, we consider only theincreases in demand for collateral over and above the steady-state level set out inSection 2.

Demand for collateral for use as initial margin (from both dealers and end-users)is assumed to increase as a linear function of financial stress (which again is proxiedby the VIX index).16 This assumption of linearity is consistent with the dynamicsof the models used by major central counterparties to calibrate their calls for initialmargin (for details see Murphy et al 2014). It might, however, overestimate the true

16 This assumption follows that in Baank for International Settlements (2013). See also Holden et al(2016).






increase in initial margin that would be called for during periods of stress, particularlyif market participants were to reduce their exposures to risk.

Increases in banks’ liquid asset buffers (held in the form of high-quality secu-rities) are also modeled as a linear function of market stress, given empirical evi-dence on banks’ liquidity hoarding behavior during the financial crisis (see, forexample, Berrospide 2012). To calibrate changes in these liquid asset buffers weregress monthly changes in UK banks’ liquid asset buffers (expressed as a proportionof their total assets) on changes in the VIX index, to obtain a linear positive rela-tionship between the two. This may be explained by banks’ propensity to increasetheir holdings of liquid assets as a precaution during periods of stress. We furtherassume that the relationship calibrated for UK banks holds for the wider bankingsystem.

3.3 Summary: the risk of demand exceeding supply (risk 1)

A summary of these drivers of changes in collateral supply and demand, and theirmodeled dependence on changes in market stress and/or CDS premiums, is given inTable 5.

Drawing these components together and combining them with the modeled rela-tionship between the VIX index and dealers’ CDS premiums described above allowsus to consider the total supply of, and demand for, high-quality collateral as a functionof the VIX index. This is illustrated in Figure 6, in which the black line shows thetotal demand for high-quality collateral. The shaded areas beneath it decompose thistotal demand into its constituent parts.

� The purple area (end-user demand – secured cash investments/secured lending)illustrates demand stemming from the need for secured reinvestment of cash(ie, cash reinvestment in reverse repos against high-quality collateral by moneymarket funds, corporate and agent lenders). This constituent of demand forhigh-quality collateral is largely “static” (with the exception of cash collateralreinvestment) and varies little with the level of market stress.

� The pink area (end-user demand – IM) illustrates demand for high-qualitycollateral for margining purposes from non-dealers (ie, from end-users ofcollateral).

� The light and dark blue areas (dealer demand – IM and dealer demand – LAB)illustrate demand by dealers for collateral for use as initial margin and in banks’liquid asset buffers, respectively.

As market stress intensifies, total demand for high-quality collateral eventuallyexceeds its made-available supply (illustrated by the dashed green curve in Figure 6).






FIGURE 6 Modelled collateral supply/demand for different levels of market stress.

15 20 25 30 35 40 45 50 55 60 65 70 75 80

Risk 1

5.0

4.5

4.0

3.5

3.0

2.5

2.0

1.5

1.0

0.5

0

US

$ (t

rillio

ns)

Market stress

Dealer demand – LABDealer demand – IMEnd-user demand – IMEnd-user demand – secured cash investment/secured lendingMade available supplyTotal demand

Current level of VIX index*

�Average level of the VIX index between June and August 2015.

This is estimated to occur when the VIX index reaches a level of around 44%, whichis labelled “risk 1”. Given that the bulk of the calibration described above is based onquarterly data, it seems natural to assume that, in order for this risk to crystallize, theVIX volatility index would have to remain at (or above) this level for a similar period(ie, around three months).

While this level of the VIX index gives an (albeit rough) guide to the level of marketstress required for this risk to crystallize, its impact is harder to judge. On the one hand,it might be natural to expect that market prices might adjust to restore the equilibriumbetween supply and demand. Such an adjustment might take the form of an increasein the returns on securities lending, which would encourage beneficial owners toincrease their securities lending and redress the mismatch between collateral supplyand demand. On the other hand, it is conceivable that if the mismatch between supplyand demand were to occur very rapidly, there might be insufficient time for beneficiallenders (particularly those not very active in securities-lending markets) to undertakesecurities lending in a volume necessary to redress the imbalance.






TABLE 5 A summary of factors leading to the variation in collateral supply/demand awayfrom their levels in benign market conditions (described in Section 2) during market stress.[Table continues on next page.]

(a) Supply

Assumed to vary withCurrent ‚ …„ ƒ

size Stress (proxied CDS CalibrationFactor (US$ tn) by VIX index) premiums methodology

Madeavailablesupply

2.91 Decreases — Regression of changesin proportion ofhigh-quality securitiesavailable for loan onchanges in dealer CDSpremiums (Figure 3)

(b) Demand



Reinvestmentof cashcollateral

0.18 Decreases Decreases Empirical regression of:• changes in

proportion of cashreinvested inhigh-quality reverserepos on changes inVIX index (Figure 5)

• percentage changesin securities lentagainst cash onchanges in dealerCDS premiums(Figure 4)

Reverse repoby real moneyinvestors

1.26 Fixed —






TABLE 5 Continued.

(b) Demand (continued)



Non-high-qualitysecuritieslendingversushigh-qualitycollateral

0.18 Fixed —

Additionaldemand forinitial margin(post theimplementa-tion ofderivativesreform)

0 Increases — Assumed to vary linearlywith the VIX index(based on inference froma hypothetical centralcounterparty initialmargin models)

Additionaldemand forbanks’ liquidasset buffers

0 Increases — Regression of changesin dealers’ liquid assetsbuffers (as a proportionof their total assets) onchanges in VIX

4 THE ROLE OF DEALERS IN COLLATERAL MARKETS AND THEPROCYCLICALITY OF THEIR BEHAVIOR

The final piece of our framework considers the role played by leveraged intermediariesin mobilizing high-quality collateral: that is, between those (principally unleveraged)end-investors that supply and demand it.

In what follows, we assume these intermediaries take the form of broker-dealers,who, via a series of repo, reverse repo and securities-lending transactions, are able topass collateral between those that supply and demand it.17

17 Prospectively, such institutions may also be disintermediated via “all-to-all” electronic platformsthat directly connect suppliers and users of collateral and/or via entities not subjected to prudentialbank regulation. But, according to recent industry analysis (see, for example, International CapitalMarket Association 2015), such alternative solutions do not currently exist on a substantial scale.






4.1 Required intermediation capacity

Not every broker-dealer transacts directly with every supplier or demander of col-lateral (or, indeed, with every other dealer), as illustrated in the network shown inFigure 7. This is consistent with the finding that interbank networks typically have a“core–periphery” structure (see Langfield et al 2014). This could be due to the costsassociated with setting up and maintaining bilateral repo relationships (eg, costs oflegal agreements and counterparty credit risk assessment) or due to the establishedinterpersonal relationships between repo desks at certain financial institutions. Thismeans that every unit of collateral that passes from those that supply it to those thatuse it has to pass through a chain of intermediaries. Using a range of global dataon repo and securities-lending transactions, we estimate the average length of the“supply chain” for high-quality collateral to be around 3.9, ie, every unit of collateralpassed between those that supply it to those that demand it passes via an average of3.9 intermediaries.18 From here on, we refer to this as the “required intermediationcapacity” associated with a given level of demand for high-quality collateral.

4.2 Dealers’ intermediation capacity

There is significant evidence to suggest that dealers scaled back on their securities-financing activities during the crisis. During a downturn, leverage may rise purely inresponse to the falls in asset prices and the corresponding falls in the mark-to-marketvalue of dealer equity. This places balance sheets under pressure, forcing dealers todeleverage, unwinding repo borrowing and thereby reducing dealers’ intermediation.

But estimating the likely degree of dealer deleveraging and consequent reductionin intermediation capacity that might occur in future episodes of market stress isimpeded by having limited historical data of the necessary granularity.

Given this, we instead incorporate the effect of dealer deleveraging by drawing on atheoretical model of dealers’ choice of leverage (and hence intermediation capacity).This is based on the premise that dealers’ intermediation capacity depends on two

18 This is the total of the high-quality collateral received by dealers, ie, that used in interdealer reposand securities-lending transactions divided by the total amount of high-quality securities on loan.This is estimated on a global basis and based on a range of sources including Financial StabilityBoard (2014), International Capital MarketAssociation (2014) and International Securities LendingAssociation (2014). Our estimate differs from that of Singh (2011), in part because it includes onlyhigh-quality collateral (whereas Singh (2011) includes all collateral, including equities). It omitsrepo/reverse repos by hedge funds, because we believe that most such transactions are “relativevalue arbitrage” trades that do not facilitate collateral intermediation.






FIGURE 7 Outstanding repo transactions between banks (red) and non-banks (blue).

This network covers UK-regulated banks and investment firms only. Red dots represent dealers and blue dotsrepresent customers. Red lines represent interdealer repo transactions; blue lines those between dealers and theirclients. More connected counterparties are clustered in the center of the network. Data is as of end-2014.

factors, both a function of the level of market stress (proxied by the VIX index):

� the equity that dealers have available to carry out securities-financing transac-tions;

� the dealers’ choice of optimal leverage, or balance sheet size, for a given levelof equity.

The quantity of equity that dealers are willing to allocate to securities-financingtransactions is modeled as a decreasing function of market stress. Our intuition isthat, as market stress intensifies, dealers experience losses that reduce the value oftheir capital (including equity allocated to repo/reverse repo transactions). The dealerequity allocated to repos is estimated as the outstanding amount of securities-financingtransactions associated with the mobilization of high-quality collateral divided bythe average dealer leverage (as inferred from Federal Reserve Board (2015)). Thesensitivity of dealer equity to the level of market stress is calibrated using the empiricalrelationship between the median return on equity for major global banks and the VIXindex.

To model changes in dealers’ optimal leverage we appeal to a theoretical model ofdealer behavior. Under this model, for any given level of stress, dealers choose a levelof leverage that maximizes their shareholders’value, subject to a regulatory constrainton their minimum leverage that, if breached, leads to their default, and the claim of






FIGURE 8 Estimated required intermediation capacity necessary to meet demand forcollateral (required intermediation capacity) versus the maximum intermediation capacityof dealers for different levels of financial stress.

16

14

12

10

8

6

4

2

0

Inte

rmed

iatio

n ca

paci

ty (

US

$ tr

illio

ns)

Risk 2

Risk 1

Current level of VIX index* 5.5

5.0

4.5

4.0

3.5

3.0

2.5

2.0

1.5

Sup

py/d

eman

d (U

S$

trill

ions

)

15 20 25 30 35 40 45 50 55 60 65 70 75 80Market stress

Max intermediation capacityMade available supply Total demand

Required intermediation capacity

�Average level of the VIX index between June and August 2015.

equity holders being reduced to zero.19 As market stress – and hence asset volatility –increases, the threat of insolvency causes shareholders’ optimal choice of leverage tofall, reducing the dealers’ capacity to act as intermediaries. (Technical details of thismodel and its calibration are given in Annex 1, available online.) Throughout, thisconstraint on dealers’ leverage is assumed to match that imposed by the regulatoryminimum leverage ratio. This is assumed to be 3% (equity as a proportion of totalassets), in line with the internationally agreed Basel III standard.20

When multiplied together, the level of equity that dealers allocate to securities-financing transactions and their optimal choice of leverage provide an estimate of theamount of balance sheet that dealers, in aggregate, make available to act as intermedi-aries in collateral markets.We can also see how this varies as a function of market stress(proxied by the VIX index). This is shown by the orange line in Figure 8, alongside

19 This is in the spirit of other literature analyzing the effects of bank regulation (see Episcopos2008).20 See Basel Committee on Banking Supervision (2014). To the extent that off-balance-sheet col-lateral flows are not included in the leverage ratio measure, our model may understate the “true”leverage of the dealer.






the required intermediation capacity (purple line) that corresponds (via a fixed scalarmultiple of 3.9, described above) to a given level of demand for high-quality collateral(shown by the black line).

4.3 The risk that the required intermediation capacity necessary tomeet a given level of collateral demand exceeds dealers’maximum intermediation capacity (risk 2)

The level of market stress at which required intermediation capacity exceeds thatwhich dealers are willing to provide is where the second risk identified in this frame-work crystallizes. Note that this risk could crystallize even in the absence of risk 1.That is, even if demand for collateral does not exceed its available supply (perhapsbecause, despite market stress, the return on securities lending adjusts and increasesthe proportion of assets that are made available for loan), the dealer intermediationrequired might exceed that which dealers are willing/able to provide, which is labelledas “risk 2”. In other words, this is the risk that collateral might get “blocked” in thenetwork of dealer balance sheets. We estimate that this risk would materialize withlevels of market stress roughly commensurate with the VIX volatility index exceeding46% over a quarter.

Although risk 2 is slightly less likely to crystallize than risk 1 (since it is triggered bya slightly higher level of market stress), it could have a significant negative impact onfinancial stability. In particular, if triggered, risk 2 will likely prevent collateral fromperforming its role of supporting the functioning of markets, including the facilitationof liquidity. The consequences of this might include the sudden inability of marketparticipants to obtain the collateral they need to manage the risks associated with theirbusiness, including payment of initial margin on derivatives transactions. Risk 2 mightalso impair dealers’ ability to fund purchases of financial assets by other leveragedinvestors (eg, hedge funds), which might have implications for market functioning andliquidity. Were this to be the case, risk 2 might impair secondary market transactionsin securities that are important for financing investment in the real economy, andtherefore have negative implications for economic growth.

5 CONCLUSION

This paper identifies and quantifies a number of procyclical behaviors of marketparticipants that cause their supply of, and demand for, high-quality collateral – aswell as their willingness and/or ability to act as intermediaries in the market forsecurities-financing transactions – to decrease/increase in response to market stress.These behaviors include increased perceptions of counterparty risk (which causes areduction in securities lending) and the likelihood that key financial intermediaries






may seek to deleverage. Demand for collateral may also increase as, for example,firms hoard liquid assets, and volatile markets cause an increase in initial margin heldagainst derivatives.

These dynamics give rise to two potential risks. The first of these – demand forcollateral exceeds its supply as market stress increases – is judged to crystallize fora level of market stress commensurate with the VIX volatility index remaining at alevel of around 44% or above for around three months. Were this risk to crystallize,however, we judge that it would have only a moderate impact on financial stability,since, at least in a matter of days, the price of accessing high-quality collateral shouldadjust to restore the equilibrium between collateral supply and demand.

A second, potentially more pernicious, risk is that, as market stress increases, theintermediation capacity of dealers required to move collateral from end-suppliers toend-demanders starts to exceed that which dealers are willing/able to provide. Ouranalysis estimates that this risk might crystallize at a similar level to that describedabove, but it might have a more significant negative effect on the market’s functioning.

These conclusions are not without caveats. Most notably, this work forms a simpleframework to describe the behavior of market participants in the market for collateralbased on historical data. Given that this behavior may differ from the richer set ofbehaviors that might be witnessed in future – including given the effects of recentlyintroduced regulation – there is considerable uncertainty around the exactitude ofthe framework. The levels of market stress at which we estimate the above risks tocrystallize should therefore be viewed as broad approximations rather than preciseestimates.

There are a number of possible extensions to this framework. Most notably, policymakers could in future consider how the effects of recently introduced regulation(including the regulatory minimum leverage ratio, initial margin and minimum haircutrequirements) might affect market participant behavior, and hence the level of marketstress at which these risks may crystallize.

DECLARATION OF INTEREST

The authors report no conflicts of interest. The authors alone are responsible for thecontent and writing of the paper.

ACKNOWLEDGEMENTS

The authors are grateful to David Aikman, Nicola Anderson, David Murphy, JonathanRelleen, Manmohan Singh, Nicholas Vause and seminar participants at the IMF andthe 2nd Annual Collateral Management Summit for useful comments, conversationsand insights. The usual caveat applies.






REFERENCES

Aguiar, A., Kenett, D., Bookstaber, R., and Wipf, T. (2016). A map of collateral uses andflows. Working Paper 16, Office of Financial Research, US Department of the Treasury.

Anderson, N., Webber, L., Noss, J., Beale, D., and Crowley-Reidy, L. (2015).The resilienceof financial market liquidity. Financial Stability Paper 34, Bank of England.

Andritzky, J. (2012). Government bonds and their investors: what are the facts and do theymatter? Working Paper, International Monetary Fund.

Bank of England (2012). Financial Stability Report, Issue 31.Bank of England (2015). Financial Stability Report, Issue 38.Bank for International Settlements (2013). Macroeconomic impact assessment of OTC

derivatives regulatory reforms. Report, August, BIS (http://bit.ly/29k9Xpn).Basel Committee on Banking Supervision (2013). Basel III: the liquidity coverage ratio and

liquidity risk monitoring tools. Report, January, BIS (http://bit.ly/1e1VwjY).Basel Committee on Banking Supervision (2014). Basel III leverage ratio framework and

disclosure requirements. Report, January, BIS (http://bit.ly/1wODLeU).Berrospide, J. (2012). Liquidity hoarding and the financial crisis: an empirical evaluation.

FEDS Working Paper Series, Federal Reserve Board.Capel, J., and Levels, A. (2014). Collateral optimisation, re-use and transformation:

developments in the Dutch financial sector. DNB Occasional Studies 12(5).Cheung, B., Manning, M., and Moore, A. (2014). The effective supply of collateral in

Australia. Reserve Bank of Australia Bulletin September Quarter.Dive, M., Hodge, R., and Jones , C. (2011). Developments in the global securities lending

market. Bank of England Quarterly Bulletin 2011 Q3, 224–233.Episcopos, A. (2008). Bank capital regulation in a barrier option framework. Journal of

Banking and Finance 32(8), 1677–1686 (http://doi.org/dvqqxb).European Systemic Risk Board (2014). Securities financing transactions and the (re)use

of collateral in Europe. Occasional Paper 6, September.Federal Reserve Board (2015). Comprehensive capital analysis and review: assessment

framework and results. Report, March, FRB (http://bit.ly/1NLwerF).Financial Stability Board (2012). Securities lending and repos: market overview and

financial stability issues. Interim Report, April 27, FSB (http://bit.ly/29y6qFg).Financial Stability Board (2014). Strengthening oversight and regulation of shadow bank-

ing: regulatory framework for haircuts on noncentrally cleared securities financingtransactions. Report, October 14, FSB (http://bit.ly/29t4JTY).

Financial Stability Board (2015). FSB Chair’s letter to G20 on financial reforms: progresson the work plan for the Antalya summit. Report, October, FSB (http://bit.ly/29PLJnS).

Fuhrer, F., Guggenheim, B., and Schumacher, S. (2015). Re-use of collateral in the repomarket. Working Paper, Swiss National Bank.

Gorton, G., and Metrick, A. (2012). Securitized banking and the run on repo. Journal ofFinancial Economics 104(3), 425–451 (http://doi.org/b9kb7t).

Holden, H., Houllier, M., and Murphy, D. (2016). I want security: stylised facts about CCPcollateral and their systemic context. Bank of England.

Infante, S. (2015). Liquidity windfalls: the consequences of repo rehypothecation. Financeand Economics Discussion Series 2015-022.Board of Governors of the Federal ReserveSystem, Washington (http://doi.org/bncn).






International Capital Market Association (2014). European repo market survey 28. Report,ICMA.

International Capital Market Association (2015). The current state and future evolution ofthe European repo market. Report, ICMA.

International Securities Lending Association (2014). Securities lending market report.Report, ISLA.

Kirk, A., McAndrews, J., Sastry, P., and Weed, P. (2014). Matching collateral supply andfinancing demands in dealer banks. Federal Reserve Bank of NewYork Economic PolicyReview 20(2), 127–151.

Krishnamurthy, A., and Vissing-Jorgensen, A. (2012). The aggregate demand for treasurydebt. Journal of Political Economy 120(2), 233–267 (http://doi.org/bkzg).

Langfield, S., Liu, Z., and Ota, T. (2014). Mapping the UK interbank system. Journal ofBanking and Finance 45(Suppl.), 288–303 (http://doi.org/bkzh).

Merton, R. C. (1973). Theory of rational option pricing. Bell Journal of Economics andManagement Science 4(1), 141–183 (http://doi.org/cvh6pc).

Murphy, D., Vasios, M., and Vause, N. (2014). An investigation into the procyclicality ofrisk-based initial margin models. Financial Stability Paper 29, May, Bank of England.

Pozsar, Z. (2011). Institutional cash pools and the Triffin dilemma of the US banking system.Working Paper, International Monetary Fund.

Rey, H. (2013). Dilemma not trilemma: the global financial cycle and monetary policy inde-pendence. Jackson Hole Economic Symposium, Working Paper 21162, National Bureauof Economic Research.

Singh, M. (2011). Velocity of pledged collateral: analysis and implications. Working Paper,International Monetary Fund.






Research Paper

The effective supply of collateral in Australia

Belinda Cheung, Mark Manning and Angus Moore

Reserve Bank of Australia, 65 Martin Place, Sydney NSW 2000, Australia;emails: [email protected], [email protected], [email protected]


ABSTRACT

High-quality assets play an important role as collateral for a wide range of transactionsand activities in wholesale financial markets. Regulatory changes introduced sincethe global financial crisis are increasing the demand for high-quality assets, therebyraising concerns about possible shortages. Drawing on data from a survey of securitiesdealers’ collateral activity in the Australian market in 2014 and 2016, this paperattempts to quantify the “effective” supply of collateral assets inAustralia by applyinga measure of supply that adjusts outstanding issuance for two important featuresof the collateral market. One feature is that a large proportion of Australian high-quality assets are held by long-term investors that do not make these assets availablefor sale, loan or use in repurchase agreements. A second feature is the ability toreuse collateral assets, thereby allowing a single piece of collateral to meet multipledemands. Using this measure, the current effective supply in 2016 is estimated to bearound A$123 billion, comprising around A$99 billion of active supply that is reusedon average 1.25 times. This amount has remained largely unchanged since the 2014survey, due to broadly offsetting changes in total source collateral and the rate ofcollateral reuse.

Keywords: collateral; collateral reuse; re-hypothecation; high-quality assets; effective supply.

Corresponding author: A. Moore Print ISSN 2049-5404 j Online ISSN 2049-5412Copyright © 2016 Incisive Risk Information (IP) Limited

27




28 B. Cheung et al

1 INTRODUCTION

Fundamental changes are underway in the functioning of wholesale financial mar-kets. These are driven in part by regulatory reforms introduced since the global finan-cial crisis as well as by behavioral changes in response to lessons learned duringthe crisis. One important change has been an increased emphasis on high-qualityassets, both within the regulatory framework and in market conventions and prac-tices. A range of new regulations, such as those that require banks to maintain higherlevels of liquidity and those that promote increased collateralization in derivativesmarkets, are likely to increase market participants’ demand for high-quality assetssubstantially.

These changes have raised concerns about localized collateral shortages. The focusof most studies on collateral supply and demand to date has been the level of out-standing issuance of high-quality assets. But it is important that policymakers andmarket participants understand the extent to which some of these assets are held byinvestors that do not make them available to meet collateral demands. They also needto consider how changes in regulation and market practices could alter the way thatcollateral assets circulate in the system. Adjusting for these factors can help to deliveran estimate of the “effective” supply of collateral assets.

This paper focuses on the use of high-quality assets for collateral purposes. Itfirst introduces the role of collateral in financial markets and describes some of thechanges occurring. With particular reference to Australia, it examines current andpotential future developments in collateral use and considers the effective supply ofhigh-quality collateral assets to meet current and future demand. To better understandthe functioning of collateral markets in Australia, and to help quantify the currenteffective supply of collateral assets, the Reserve Bank ofAustralia (RBA) surveyed thelargest securities dealers in Australia on their institutions’ collateral market activity;this survey was first conducted in mid-2014, and then again in early 2016.

This paper is similar to work by the Financial Stability Board (FSB) on possiblemeasures of collateral reuse (Financial Stability Board 2016). The FSB notes theimportance of understanding the extent of collateral reuse, given its potentially ben-eficial role in alleviating collateral scarcity as well as its potential risks to financialstability. Currently, data on reuse is not widely available. It is typically collected inmore ad hoc surveys, such as the one used in this paper. The FSB (2016) discusses anumber of data elements that could be included in its global securities financing datastandards, many of which are similar to the data collected in the surveys reported inthis paper. The FSB’s Data Experts Group is currently developing recommendationson potential measures of collateral reuse and on related data requirements; it plans toreport by the end of 2016.





The effective supply of collateral in Australia 29

2 COLLATERAL USE IN WHOLESALE MARKETS

The basic role of collateral is to manage counterparty credit risk. Several typicalattributes of collateral make it an effective tool for doing so. Collateral provides reli-able and timely protection in the event of a default and provides a senior claim inbankruptcy. Compared with an unsecured exposure, collateral alleviates the informa-tion asymmetry between the borrower and lender regarding the borrower’s credit-worthiness, because the lender’s principal interest is in the quality of the collateral.Further, collateral helps to align the incentives of borrowers and lenders: unsecuredborrowers may have an incentive to take riskier decisions, since the risk is ultimatelyborne by the lender; secured borrowers, by contrast, risk losing their collateral.

Accordingly, collateral assets in wholesale markets are typically of high quality –that is, assets with low credit, market and liquidity risks – so they would be expectedto retain their value and could be liquidated on a timely basis, should the counterpartydefault. For instance, repurchase agreements (repos), the most common form of col-lateralized lending in wholesale markets, are typically contracted against a defined setof high-quality assets. In Australia, most repos are contracted against “general collat-eral”, which includes Australian Government securities (AGS) and securities issuedby the states and territories (“semi-government” securities). Currently, around 90% ofoutstanding repos are backed by government-related securities, and most repos havematurities of less than fourteen days.

The repo market is the most significant venue for the exchange and circulationof high-quality assets in the domestic financial market, and it plays an importantrole in institutions’ funding and liquidity management activities. Active participantsin the domestic repo market include securities dealers – typically large domesticand international banks that are market makers in government securities – as well assome smaller institutional non-dealer participants and the RBA (Wakeling and Wilson2010). Two of the most significant areas of repo market activity in Australia are thefollowing.

RBA open market operations. Repos offer a flexible instrument for the RBA to man-age the total amount of outstanding Exchange Settlement Account (ESA) balancesin the banking system, so as to keep the cash rate as close as possible to the tar-get set by the Reserve Bank Board. By executing repos with its counterparties,principally as a cash provider, the RBA manages the aggregate of institutions’ESAbalances. As of February 2016, repos with the RBA accounted for around one-thirdof outstanding repo market positions.1

1 This excludes banks’“open repos” with the RBA for the purpose of meeting settlement obligations.





30 B. Cheung et al

Market making in government securities. Securities dealers are major participants inthe Australian repo market. As market makers in domestic government securities,dealers match buyers and sellers of the same security, or – when timing mismatchesarise – buy and sell for their own account. The repo market facilitates this activity.In particular, dealers are able to fund their inventory of securities by selling themunder repo. Selling securities under repo allows dealers to raise funding withouthaving to liquidate outright positions; alternatively, a dealer can source fundinginternally from its treasury desk. The principal provider of cash to the Australianrepo market is the RBA. Dealers may also use repos to borrow securities they haveagreed to sell to their customers. Much of this activity – around 20% of outstandingrepo positions – occurs between securities dealers. Investment funds and othernon-dealer institutions are also providers of securities to securities dealers. Theseinstitutions typically use repos to manage their short-term funding without sellingtheir high-quality assets outright.

Securities lending activity also supports the circulation of high-quality assets inwholesale markets. While in many ways similar to repo activity, securities lendingis typically driven by the need to hold a particular security – often to meet a marginrequirement or to cover a short sale or a failed settlement. Loaned securities areusually sourced from investment funds or superannuation funds. These funds typicallyoperate via custodian banks that act as securities lending agents. In the Australiansecurities lending market, most loaned securities are equities, with only around one-third of securities loans involving fixed income securities (Markit 2013). Loans maybe collateralized by cash or other non-cash assets (subject to a haircut). Securitieslending agents then reinvest the cash collateral received.

Clearing via central counterparties (CCPs) is another source of collateral demand.CCPs help to manage counterparty credit risk in a wide range of markets, includingequity, fixed income and, increasingly, over-the-counter (OTC) derivatives markets(see Section 3).2 To manage the financial exposure it assumes in carrying out itsfunction, a CCP collects collateral from its participants: initial margin, to managepotential future price changes before an exposure to a defaulted participant’s positioncan be closed out; and variation margin, to cover observed changes in the mark-to-market value of participants’ open positions.3 CCPs also typically collect collateral

2 A CCP stands between the buyer and seller in a financial market transaction. The CCP guaranteesthat if one party was to default on its obligations to the CCP, the CCP would continue to meet itsobligations to the other.3 CCPs collect initial margin from both participants to a given position, but variation margin iscollected only from the participant with a mark-to-market loss and is passed to the other participant(who has an equivalent mark-to-market gain).






from participants to fund a buffer of pooled financial resources in case a defaultedparticipant’s margin proves insufficient.

The collateral used to meet CCPs’margin requirements may take the form of cash ornon-cash assets. Variation margin is generally always met in cash, since it is typicallypassed through from the participant with a mark-to-market loss to the participant witha gain. Initial margin requirements, however, may be met using cash or high-qualityassets.4 In Australia, for instance, cash is commonly posted to meet initial marginrequirements at the two domestic CCPs (ASX Clear and ASX Clear (Futures)).5 Asof the end of December 2015, on average, a little more than half of initial marginobligations at ASX Clear were met with cash; at ASX Clear (Futures), around 97%of initial margin requirements were met with cash.

Collateral can also be exchanged between counterparties to noncentrally clearedOTC derivative contracts. To date, this has typically involved only the exchange ofvariation margin. Initial margin has not been widespread, although this is changing(see Section 3).

3 THE INCREASING DEMAND FOR HIGH-QUALITY ASSETS

Non-regulatory demand for high-quality assets has been increasing as investors haveshifted toward more collateralized lending. However, recent and upcoming regulatorychanges are also driving an increase in the demand for high-quality assets for bothcollateral and non-collateral purposes. Given these regulatory changes, there havebeen a number of studies into how financial markets might respond and what changesmight be needed to ensure the efficient mobilization of collateral to meet increaseddemand. For instance, Heath and Manning (2012) discuss high-quality liquid assets(HQLAs) in Australia; Fender and Lewrick (2013) discuss the implications for theglobal market for collateral as well as possible adjustment mechanisms in marketswhere HQLAs are more scarce; and the European Central Bank (2013, 2014) andMersch (2014) consider the differing collateral-eligibility standards across the Euro-pean Union and recommend changes to ensure availability and efficient mobilizationof HQLAs.

4 International standards introduced in 2012 clarify requirements around the size and composition ofCCPs’ pooled financial resources, and also set expectations around eligible non-cash collateral andthe reinvestment of cash collateral (Committee on Payment and Settlement Systems and TechnicalCommittee of the International Organization of Securities Commissions 2012). The AustralianSecurities and Investments Commission (ASIC) and RBA have implemented these standards inAustralia (Australian Securities and Investments Commission and Reserve Bank of Australia 2013).5 ASX Clear provides CCP services for equities and equity options; ASX Clear (Futures) clearsexchange-traded futures and OTC derivatives.





32 B. Cheung et al

3.1 Increased demand for collateral purposes

The most significant regulatory changes relate to the way counterparty risk is managedin the OTC derivatives market. While central clearing has long been a source ofcollateral demand, the range of products covered by CCPs’ activities is expanding.Since the global financial crisis, the move to central clearing of OTC derivatives hasaccelerated following the Group of Twenty’s (G20’s) commitment in 2009 to ensurethat all standardized OTC derivatives are centrally cleared. Mandatory central clearingis already in place in a number of jurisdictions – as of November 2015, twelve of thetwenty-four FSB jurisdictions had frameworks in place (Financial Stability Board2015).6 Further, under the Basel bank capital regime, there are incentives to centrallyclear derivative positions. By February 2014, around three-quarters of the outstandingvalue of interest rate derivatives globally (the largest segment of the OTC derivativesmarket) had therefore been centrally cleared (International Swaps and DerivativesAssociation 2014).

Since noncentrally cleared trades typically did not previously involve either sideposting initial margin, the transition to central clearing has been accompanied by anincrease in the demand for, and use of, high-quality assets.7

The proportion of noncentrally cleared derivatives transactions that is collateral-ized, at least with variation margin, has also increased significantly over the pastdecade. A recent survey by the International Swaps and Derivatives Association(ISDA) reports that, globally, around 90% of noncentrally cleared transactions incredit, fixed income and equity derivatives, and around three-quarters of foreignexchange derivatives, are subject to a collateral agreement (International Swaps andDerivatives Association 2015).

From September 2016, collecting both variation and initial margin on noncentrallycleared transactions will become a mandatory requirement globally for transactionsbetween certain counterparties (Basel Committee on Banking Supervision and theInternational Organization of Securities Commissions 2015).

3.2 Increased demand for non-collateral purposes

At the same time, regulatory changes are increasing the demand for high-qualityassets for purposes other than collateral requirements. In particular, the liquidity

6 In Australia, in response to recommendations from the Australian Prudential Regulation Authority(APRA), ASIC and RBA, the government has adopted mandatory clearing for interest rate deriva-tives denominated in the major currencies and the Australian dollar; these requirements came intoforce in early 2016 (Australian Securities and Investments Commission 2015).7 In addition, the international standards for CCPs require segregation of client assets. These addi-tional protections have the effect of increasing collateral demand, since they reduce the scope fornetting against client positions.






coverage ratio (LCR) introduced under the Basel III reforms requires that bankshold an amount of specified high-quality assets sufficient to withstand thirty days ofoutflows in stressed market conditions.8

In the Australian context, APRA has defined these HQLAs to comprise reservebalances with the RBA, AGS and semi-government securities. Banks have adjustedtheir high-quality asset holdings in response to the regulations formally taking effectin 2015. The share of HQLAs on Australian banks’ balance sheets has risen to morethan 11% of banks’ total assets since the global financial crisis, with the proportionof these held in AGS and semi-government securities increasing from just 10% toalmost 55%.9

3.3 Estimates of increased demand for high-quality assets

There have been a number of attempts to quantify the implications of some of theseregulatory developments for the demand for high-quality assets in a range of marketsglobally. Estimates of the implications of central clearing and margining of noncen-trally cleared derivatives are sensitive to assumptions about the volatility of clearedproducts, the proportion of OTC derivatives trades that will eventually transition tocentral clearing, and the extent to which trades will be fragmented across multipleCCPs (Table 1). Several studies have emphasized the greater scope for collateralefficiency if trades are centrally rather than noncentrally cleared (Heath et al 2013).These studies also highlight the efficiencies of concentrating clearing in one CCP ora few CCPs, rather than clearing in several CCPs operating in different markets orproducts (Duffie and Zhu 2011).

4 THE EFFECTIVE SUPPLY OF HIGH-QUALITY COLLATERALASSETS

While the range of estimates is quite wide, the studies in Table 1 have typicallyconcluded that a global shortage of high-quality assets is unlikely. The Committee onthe Global Financial System (2013), for instance, notes that while demand for high-quality assets could increase by an estimated US$4 trillion as a result of regulatorychanges, between 2007 and 2012 the supply of high-quality government securitiesincreased by US$10.8 trillion.

8 Foreign bank branches and smaller authorized deposit-taking institutions (ADIs) face differentLCR requirements to larger domestically incorporated ADIs; the former institutions have a 40%coverage requirement instead of the 100% requirement applied to larger domestically incorporatedADIs. This serves to somewhat lessen the demand for Australian HQLAs by these foreign bankbranches and smaller ADIs.9 This is based on data for December 2015 from APRA and the RBA.





34 B. Cheung et al

TAB

LE

1E

stim

ates

ofin

tern

atio

nald

eman

dfo

rhi

gh-q

ualit

yas

sets

.

So

urc

eC

over

age

Ran

ge

of

esti

mat

es

Duf

fieet

al(2

015)

Sub

seto

fthe

OT

Ccr

edit

deriv

ativ

esm

arke

t:in

itial

mar

gin.

4.5–

8%of

netn

otio

nalo

utst

andi

ng,d

epen

ding

onth

em

arke

tstr

uctu

reus

ed.A

num

ber

ofal

tern

ativ

esar

eco

nsid

ered

.C

GF

S(2

013)

Glo

bal:

LCR

;ini

tialm

argi

nfo

r(c

entr

ally

and

nonc

entr

ally

clea

red)

OT

Cde

rivat

ives

.E

stim

ated

incr

ease

ofU

S$4

trill

ion.

Cru

zLo

pez

etal

(201

3)G

loba

l:O

TC

deriv

ativ

es,i

nclu

ding

uncl

eare

dm

argi

ning

and

limits

tore

-hyp

othe

catio

n;LC

R.

US

$2to

US

$4tr

illio

nby

the

end

ofth

eph

ase-

inof

all

requ

irem

ents

.

Cap

elan

dLe

vels

(201

2)E

uro

area

:LC

R;i

nitia

land

varia

tion

mar

gin

for

(cen

tral

lyan

dno

ncen

tral

lycl

eare

d)O

TC

and

exch

ange

-tra

ded

deriv

ativ

es;E

uros

yste

mop

erat

ions

;rep

om

arke

tact

ivity

.

€4.

7tr

illio

nby

the

end

of20

14.T

his

refle

cts

anes

timat

edin

crea

seof

€2

trill

ion

betw

een

2012

and

2014

.

Hel

ler

and

Vau

se(2

012)

Larg

estf

ourt

een

glob

alde

aler

s:in

itial

mar

gin

for

cent

rally

clea

red

OT

Cin

tere

stra

tean

dcr

edit

deriv

ativ

es.

Sep

arat

eC

CP

for

each

asse

tcla

ss(h

igh-

vola

tility

scen

ario

):U

S$1

07bi

llion

for

cred

itde

rivat

ives

;US

$43

billi

onfo

rin

tere

stra

tede

rivat

ives

.The

estim

ates

vary

sign

ifica

ntly

acco

rdin

gto

the

chos

envo

latil

ityle

vel.

ISD

A(2

012)

Glo

bal:

initi

alm

argi

nfo

rno

ncen

tral

lycl

eare

dO

TC

deriv

ativ

esac

ross

alla

sset

clas

ses.

US

$800

billi

on–U

S$1

.7tr

illio

nin

norm

alm

arke

tco

nditi

ons,

with

firm

sus

ing

inte

rnal

mar

gin

mod

els

toca

lcul

ate

requ

irem

ents

.

Sid

aniu

san

dZ

ikes

(201

2)G

loba

l:in

itial

mar

gin

for

cent

rally

and

nonc

entr

ally

clea

red

OT

Cin

tere

stra

tean

dcr

edit

deriv

ativ

es.

Bet

wee

nU

S$2

00bi

llion

and

US

$800

billi

on,d

epen

ding

onth

ene

tting

effic

ienc

yac

hiev

edby

cent

ralc

lear

ing.

Thi

sre

flect

san

estim

ated

incr

ease

ofU

S$1

30bi

llion

toU

S$4

50bi

llion

from

pre-

refo

rmle

vels

ofce

ntra

lcle

arin

g.

CG

FS

isth

eC

omm

ittee

onth

eG

loba

lFin

anci

alS

yste

m.






Nonetheless, it is acknowledged that while the total supply of high-quality assets isimportant, the geographical distribution of that supply also matters. Localized short-ages could arise. This may be particularly important for markets, such as Australia,that have a smaller supply of government debt outstanding. Further, looking solelyat the supply on issue will not fully capture the availability of high-quality assets tomeet collateral needs.

The “effective supply” of collateral is more indicative of both the availabilityof high-quality assets to support collateral-dependent activities and the way theseassets are used. Determining effective supply requires two important adjustments toa measure of the total supply of high-quality assets, with partially offsetting effects.

Active supply. This is the supply of high-quality assets after adjusting for assets thatare “locked away” in buy-and-hold portfolios and are thus unavailable for sale, loanor repo. They may alternatively be unavailable because they are held to meet cer-tain minimum regulatory requirements. These assets may therefore be considered“inactive” for collateral purposes.

Collateral reuse. In some collateralized transactions, the collateral receiver has thelegal right to reuse the collateral. Reuse allows a single piece of collateral to simul-taneously support multiple demands and assists in intermediation between sourceproviders of collateral assets and the ultimate users of those assets. The sourceprovider of a collateral asset may be thought of as the starting point in a “collat-eral chain”, with the ultimate user being the end point. For instance, CCPs do nottypically reuse collateral received, other than in exceptional circumstances, suchas the default of a clearing participant. While collateral reuse helps a participant ina CCP to access the collateral that it needs to meet the CCP’s margin requirements(through, for instance, repo markets), the delivery of the collateral to the CCP isthe end point in the collateral chain. Singh (2013) describes the important roleof collateral reuse, which he terms “collateral velocity”, in supporting wholesalefinancial market activity. Securities dealers have traditionally relied significantly ontheir ability to reuse collateral received under repo from institutional investors for arange of activities, including the following: supporting their market-making activityin the government bond market; raising short-term funding and managing short-term liquidity needs; meeting other market participants’ demand for high-qualityassets; and matching repo or derivatives trades between clients.

Taking into account these two adjustments, the effective supply of collateral maybe calculated by first subtracting the inactive component from the total supply ofhigh-quality assets on issue to yield the “active supply”, and then multiplying thisactive supply by an estimate of the number of times that each piece of collateral is





36 B. Cheung et al

reused on average:

active supply � reuse of collateral D effective supply;

where

reuse of collateral D total collateral use

total source collateral:

Of course, even if there is a sufficient effective supply of high-quality assets, access-ing these assets may be more challenging or more costly for some institutions. Forinstance, where nonfinancial corporations or investors use derivatives markets only tohedge illiquid assets or future cashflows, increased collateralization could be a sourceof liquidity risk (Australian Prudential Regulatory Authority, Australian Securitiesand Investments Commission and Reserve Bank of Australia 2014).

Our measure of collateral reuse is similar to others used in the literature. Fuhrer et al(2015) describe an identical measure of collateral reuse as the “collateral multiplier”.Similarly, the FSB (2016) sets out a number of potential measures of collateral reuse.Ours is similar to the proposed “reuse rate” measure. However, since we include totaluse in the numerator, rather than isolating reuse activity, our measure is equivalent toone plus the FSB’s proposed reuse rate measure.

5 THE EFFECTIVE SUPPLY OF COLLATERAL IN AUSTRALIA

Applying the concepts introduced above, this section attempts to estimate the effectivesupply of high-quality assets to meet collateral demand in Australia. This analy-sis focuses on the highest-quality collateral issued in Australia: AGS and semi-government securities. As noted, these assets are currently the most commonly usedform of noncash collateral in Australia. While it may be argued that some institutionsmanage liquidity on a global (or regional) basis, this paper takes the view that ananalysis of effective supply at the domestic level is most appropriate in the Australiancontext. This reflects that domestic banks are held to a liquidity standard specifiedin terms of Australian HQLAs, and many collateral obligations may be met usingonly Australian dollar-denominated assets – including access to liquidity facilitiesprovided by the RBA.

This section begins with a discussion of the holders of Australian high-qualityassets. This shows that a large proportion of Australian high-quality assets are heldfor non-collateral purposes and are not made available for collateral purposes. Theyare therefore not part of the “active supply”. It then introduces data from a surveyof Australian securities dealers’ collateral market activity to help estimate the activesupply of AGS and semi-government securities and the rate of collateral reuse.






FIGURE 1 Holders of federal government debt: proportion of securities outstanding.

20092003 2015 20092003 20150

10

20

30

RBA% %

60

40

20

0

Domestic pension andinsurance funds

Nonresidents

Domesticbanks*

Otherdomesticfinancial**

* Includes other domestic depository institutions. ** Includes domestic investment funds. Source: ABS.

5.1 Holders of Australian high-quality assets

The majority of Australian high-quality assets are held by nonresident entities anddomestic banks. The proportion of total outstanding AGS that is held by nonresidententities has grown substantially over the past two decades, to a little less than two-thirds, or almost A$280 billion of the more than A$450 billion on issue, as of the endof December 2015 (see Figure 1).

Around a quarter of total issuance of semi-government securities, or about A$70billion, is also held by nonresident entities (Figure 2). These entities typically donot use these assets for collateral-related activities and generally do not make themavailable for such activities. As a result, a very large proportion of Australian high-quality assets are inactive for collateral purposes.10

As noted, domestic banks’ holdings of high-quality assets rose markedly in antici-pation of the liquidity regulations introduced in 2015; banks’ holdings of AGS haveincreased nearly twentyfold from June 2008 – just prior to the height of the globalfinancial crisis – to around A$70 billion, and their holdings of semi-governmentsecurities nearly sevenfold to more than A$150 billion. Given the dominance of non-resident investors and domestic banks, other domestic private sector investors (such

10 The RBA does not itself reuse collateral; however, it allows participants to freely switch thesecurities used as collateral. Thus, these securities are not entirely “locked away”.





38 B. Cheung et al

FIGURE 2 Holders of state and local government debt: proportion of securities outstand-ing.

20092003 20150

20

40

%

20092003 20150

20

40

%

RBA

Domestic pension andinsurance funds

Nonresidents

Domesticbanks*

Otherdomesticfinancial**

*Includes other domestic depository institutions. **Includes domestic investment funds. Source: ABS.

as pension, insurance and investment funds) collectively hold only around a tenth oftotal issuance of AGS and around a fifth of semi-government debt.

5.2 A survey of Australian securities dealers’ collateral use

To better gauge the effect of the developments described in this paper on the function-ing of the Australian collateral market, and to assist in estimating both active supplyand the rate of collateral reuse, the RBA surveyed the nineteen largest securities deal-ers in Australia on their collateral market activity as of June 2014 and March 2016.A particular focus was collateral reuse.11

To the extent that most collateral market activity involving Australian high-qualityassets is intermediated by major securities dealers, a survey of these entities shouldprovide a reasonable estimate of collateral use and reuse in the Australian market as a

11 One limitation of the survey is that it can identify only reuse of collateral received from thirdparties; it cannot identify reuse of collateral within an institution, eg, between an institution’s treasuryfunction and its repo desk.






TABLE 2 Collateral use and reuse by securities dealers in the Australian market(a).

Of which:(e)

received asOwned outright collateral

and pledged/ Received as and pledged/repo’d/loaned to collateral from repo’d/loaned to‚ …„ ƒ ‚ …„ ƒ ‚ …„ ƒ2014 2016 2014 2016 2014 2016

Other survey respondents 12.8 13.5 20.7 19.1 7.5 6.7Institutional investors 2.9 1.4 17.1 24.9 1.9 0.6Other banks 1.7 3.0 11.7 8.9 3.0 0.7Central banks(c) 17.5 22.3 0.3 2.3 24.6 10.6CCPs 0 2.1 0 0 0 0.5

Total(d) 35.0 42.3 49.8 55.3 37.1 19.1

Outstanding positions, A$ billion, GC 1 securities,(b) June 2014 and March 2016.(a) Based on a survey of nineteen securities dealers in the Australian repo market. Due to non-response by somesmaller entities, the reported numbers do not capture all collateral activity; as a result, the collateral received bysurvey respondents from other survey respondents does not balance exactly the collateral pledged/repo’d/loanedby survey respondents to other survey respondents.(b) Includes actively traded treasury bonds and semi-government bonds, treasury notes and Australian and stategovernment indexed bonds; note that GC 1 assets are a subset of APRA defined HQLAs, in that only actively tradedAGS and semi-government securities are eligible for GC 1.(c) Includes the RBA and, to a small extent, foreign central banks.(d) Components may not sum to totals due to rounding.(e) Of total collateral received.Source: RBA.

whole.12 The survey sought a breakdown of dealers’counterparties and also separatelyidentified dealers’ activities using the highest quality general collateral (GC 1) (seeTable 2).

As Table 2 shows, a material proportion of the collateral used by dealers drawson assets that these institutions own outright. Indeed, this proportion was more thantwo-thirds in 2016. Consistent with data cited elsewhere (Wakeling and Wilson 2010),Table 2 confirms the relatively high level of activity between securities dealers.Also, asexpected, more than half of the collateral received by respondent securities dealers isprovided by institutional investors and other banks, which includes securities lendingagents acting on behalf of institutional investors.

Table 2 also emphasizes the important role of the RBA as a collateral receiver.Survey respondents used aroundA$42 billion inAGS and semi-government securitiesin June 2014 to support their participation in the RBA’s operations, with around 60%

12 Non-intermediated collateral activities, such as direct securities lending between a foreign institu-tional investor and a hedge fund, are not captured by the survey. This is because non-intermediatedcollateral activities are uncommon in the Australian market. Thus, our estimated collateral reuserate represents a lower bound.





40 B. Cheung et al

of this amount comprising reused securities.13 While collateral repo’d to the RBAremained a high proportion of overall collateral use by survey respondents in March2016, the amount outstanding with the RBA was somewhat lower, at A$33 billion.Further, only a third comprised reused securities.

Also notable is the increase in collateral used to support clearing activities at CCPs.Respondents reported no use of collateral for this purpose in 2014, but they reportedA$2.6 billion pledged or repo’d to CCPs in 2016, of which A$0.5 billion comprisedreused securities.

The survey additionally sought information on collateral activities using the broadersecond tier of general collateral (GC 2), which includes some nongovernment-relatedsecurities. By comparison with the data on GC 1 assets, the use of GC 2 assets ascollateral is much lower: around A$32 billion in total, compared with A$98 billionin GC 1 assets. Of this, around A$5 billion was used by securities dealers to accesscentral bank operations.

5.3 Active supply of high-quality assets in Australia

Adjusting for the inactive component of the stock of high-quality assets may beparticularly significant in the Australian context. This is because a large proportion ofhigh-quality assets on issue is held overseas and understood to be held by long-terminvestors – largely official sector investors – that typically do not make their assetsavailable for sale, loan or repo.

The survey data provides a basis for estimating how much of the outstanding supplyof AGS and semi-government securities is currently “actively used” as collateral –either under pledge, repo or a securities lending agreement. This is outlined in Table 3.

According to the 2016 survey data, banks and securities dealers currently activelyuse around A$51 billion of AGS and semi-government securities that they own out-right, up from A$47 billion in 2014. Institutional investors, such as superannuationor insurance funds, provide around A$24 billion of high-quality assets for collateralpurposes; this also represents an increase from A$17 billion in 2014. The currentactively used supply is therefore around A$77 billion, which, while higher than theA$64 billion observed in 2014, is low relative to the outstanding supply of these assets.

The actual available supply of high-quality assets is, however, greater than thatwhich is currently actively used. This includes securities committed to securitieslending programs that are not currently on loan. The utilization rate of governmentsecurities committed to lending programs is only around one-third (Figure 3), leavingapproximately A$21 billion committed, but unutilized. These securities should alsobe regarded as part of active supply.

13 The sum of the first and third major columns of the “central banks” row of Table 2 (for 2014 and2016, respectively).






TABLE 3 Holdings and use of Australian government debt.

Holdings:(a)

% of total ActivelyHoldings:(a) outstanding used(b)

A$ billion securities (A$ billion)‚ …„ ƒ ‚ …„ ƒ ‚ …„ ƒ2014 2016 2014 2016 2014 2016

Banks and securities dealers 168.4 224.7 26.6 30.4 46.7 51.2Institutional investors 102.9 118.4 16.3 16.0

17.1 24.9Nonresidents 318.2 352.9 50.3 47.7Other(c) 43.0 43.9 6.8 5.9 0.3 2.3

Total(d) 632.5 739.9 100.0 100.0 64.1 78.4

“Holdings” data as of June 2014 and December 2015 and “actively used” data as of June 2014 and March 2016.(a) Holdings refers to the securities held by the institution as of the reporting date. To the extent that securities havebeen pledged/repo’d/loaned to the institution, these would be included in the holdings figures. Accordingly, this datadoes not capture the sources of actively used high-quality assets but rather the final end points of collateral chains.(b) The data is drawn from data in Table 2 on securities owned outright and pledged/repo’d/loaned by respondentsecurities dealers, and securities received as collateral by respondent securities dealers (other than from otherrespondent securities dealers). Note that the breakdown of institution types in the survey data does not matchprecisely the breakdown in the data on holders of high-quality assets. Also, data on active use by institutionalinvestors may be higher, and those for banks and securities dealers correspondingly lower, to the extent that use bybanks and securities dealers captures intermediation of collateral use by institutional investors (eg, banks acting assecurities lending agents for institutional investors).(c) Includes RBA, federal, state and local government, and public and private nonfinancial corporations.(d) Components may not sum to totals due to rounding.Sources: ABS; Data Explorers; Markit; RBA.

Adding the unutilized component of securities committed to lending programs tothe A$78 billion of actively used supply in March 2016 yields a total active supplyof around A$99 billion, or around 13% of total AGS and semi-government securitiesoutstanding. While active supply is more than 20% higher than at the time of the2014 survey, as a proportion of total government securities outstanding it is largelyunchanged.

5.4 Collateral reuse and effective supply in Australia

Table 2 may be used to estimate the current rate of collateral reuse. The relevantmetrics are the following.

Total source collateral. This is equivalent to the current actively used supply, calcu-lated above as around A$78 billion.

Total collateral use. This may be estimated from the sum of total source collateral(above) and total collateral received and then reused by respondent dealers (ie, the“Total” in the third major column of Table 2): around A$98 billion.





42 B. Cheung et al

FIGURE 3 Securities lending in government debt: amounts outstanding.

20142012 20160

5

10

15

20

$b

20142012 20160

5

10

15

20

$b

Australian governmentsecurities

Semi-governmentsecurities

Available forlending

On loan

Sources: Data Explorers; Markit.

The rate of collateral reuse may then be estimated by dividing total collateral useby total source collateral. This returns a rate of current reuse of approximately 1.25times, down from 1.6 times in 2014.

Applying the rate of collateral reuse to the estimate of active supply derived earlier,A$99 billion, yields an effective supply of A$123 billion. With broadly offsettingchanges in total source collateral and the rate of collateral reuse, the effective supplyin 2016 is similar to the A$128 billion calculated in 2014. This is around one-sixthof total outstanding issuance.

6 CHANGES IN THE EFFECTIVE SUPPLY OF COLLATERAL

Importantly, neither active supply nor reuse are fixed quantities. The regulatorychanges identified above will have implications not only for demand, but also for theway that collateral circulates through the system and is ultimately used. Increased useof CCPs and greater segregation of derivatives-related client margin, for instance, willreduce the rate of reuse of collateral. More generally, market participants’ responsesto changes in relative prices, loan terms and repo rates will influence their investmentand asset allocation decisions, potentially altering both active supply and the rate ofreuse.






Even between the two surveys considered in this paper, there have been materialchanges in both active supply and reuse, with active supply up by more than 20% andthe rate of reuse down from 1.6 to 1.25. The increase in active supply would seemto reflect a 17% increase in total outstanding issuance of government securities andan increase in the share of outstanding issuance held by banks and securities dealers,relative to nonresidents. The decline in the rate of collateral reuse is more difficultto explain, since it is calculated at only a single point in time. A number of factorsmay be at play, some of which are discussed below, but in the absence of a longertime series it is not possible to draw definitive conclusions on either the trend or thedrivers of the observed changes. The authors’ liaison with survey respondents has notrevealed a material change in behavior in the short period between the two surveys.

6.1 Potential changes in active supply

Active supply could, of course, change as the total issuance of high-quality securitieschanges, depending on how, and by which types of institutions, new net supply isabsorbed into the market. Over time, adjustments in relative asset prices could alsocreate an incentive for some existing holders of high-quality assets to reallocate theirportfolios. However, there is traditionally “stickiness” in many investment mandates.Even price-sensitive investors may adjust with a considerable lag. Further, someinvestors will naturally react slowly to price changes, perhaps because their fixedincome investments are hedging long-term cashflows or other liabilities. Nevertheless,if loan terms or repo rates on high-quality assets became attractive, some of these assetscould be encouraged into securities lending programs or repo markets, and therebybecome “active”.

Although it is not possible to reliably predict how active supply will change overtime in response to changing relative prices or changing loan terms and repo rates,some observations may be made.

� In the case of banks and securities dealers, there may be limited scope foradditional government-related asset holdings to be actively used as collateral.As noted, the increase in these institutions’ holdings is largely in response tothe formal introduction of the LCR. Accordingly, to the extent that relativeprices encouraged banks to retain these holdings in government-related secu-rities rather than ESA balances, they would need to remain unencumbered onbanks’ balance sheets. In 2016, APRA estimated that the LCR could generatea total demand for HQLAs in excess of A$400 billion. If met entirely withgovernment-related securities, this would require a substantial further increasein banks’ holdings of these securities to more than 50% of the total outstand-ing supply. Such an increase is unlikely to be possible. Even if it were pos-sible, it would cause significant disruption to the functioning of the market





44 B. Cheung et al

and adversely affect liquidity. Hence, it would be self-defeating. Given this,authorized deposit-taking institutions (ADIs) are able to establish a committedliquidity facility (CLF) with the RBA to help meet these requirements. With theCLF, the RBA commits to making available a pre-agreed amount of liquidityunder repo against any securities that are eligible in the RBA’s normal opera-tions. This extends beyond AGS and semi-government securities.14 A numberof institutions have established a CLF, and more than half of LCR-requiredliquid assets are currently being met with the CLF; self-securitized residentialmortgage-backed securities comprise the bulk of CLF collateral, thus easingthe demand for government securities (Debelle 2015).A related justification forthe CLF is that, if a supply–demand imbalance left government securities tooexpensive, banks would meet the LCR using ESA balances with the RBA. Thiswould have the potential to affect monetary policy implementation by makingthe demand for ESA balances unstable (Debelle 2011).

� Given the current very low active use of AGS and semi-government securitiesby institutional investors/nonresidents as a proportion of their holdings, therelease for loan or repo of a small additional proportion of these holdings couldmaterially increase the active supply of these securities. Indeed, as noted, therewas some evidence of this between 2014 and 2016.

6.2 Potential changes in collateral reuse

Similarly, the rate of collateral reuse is not fixed. Faced with tightness in collateralmarkets, financial institutions could seek to reuse the collateral they receive moreeffectively. This may be supported by the emergence of centralized collateral man-agement services and other technological or institutional advances that assist marketparticipants in optimizing their collateral use.15

However, the scope to increase the rate of collateral reuse may at the same time belimited by regulatory and behavioral developments. As noted, increased use of CCPs(and segregation of client assets at CCPs) will reduce collateral reuse, since CCPsare end points in collateral chains. Indeed, between 2014 and 2016, CCPs emergedas a significant end point, with survey respondents reporting A$2.6 billion in GC 1collateral allocated to CCPs in 2016, relative to none in 2014.

14 ADIs are charged a fee of 15 basis points for this commitment. The fee has been set to reflect thetypical liquidity premium between assets eligible for the CLF and HQLAs, thereby leaving ADIsindifferent between using HQLAs and the CLF.15 One implication of increased use of triparty services is that the institution facing a collateraldemand may not directly control the collateral selection process, so it may not make a strategicdecision as to whether to use outright-owned securities or reuse collateral received. This may makeit more difficult to interpret changes in the calculated rate of reuse.






Restrictions on reuse in the forthcoming international standards for margining ofnoncentrally cleared derivatives could have a similar effect.

In addition, behavioral changes are important. With the experience of the default ofLehman Brothers, there is now increased awareness of the risks of re-hypothecationand reuse of collateral. Some investors no longer permit the practice; others arerestricting it and are requiring greater transparency of the activity. In a similar vein,investors are increasingly seeking better segregation of client assets and managingexposures to their agents more carefully.

The effect of these developments has been observed in other markets. Singh (2013)reports that collateral reuse globally has fallen sharply since the global financialcrisis. This has coincided with an observed decline in repo activity. While this may, inpart, reflect the factors discussed above, it also likely reflects more stringent leverageratio requirements that have been applied in key securities financing markets. Lowerleverage would tend to lower the rate of collateral reuse.

7 DISCUSSION

At present, the current effective supply of high-quality assets would seem to be suf-ficient to support collateral demand. While there was some increase in repo ratesbetween 2014 and 2016, there is no material evidence of a shortage of securitiesavailable for repo. Further, the utilization rate of government-related securities com-mitted to lending programs is relatively low. The Australian collateral market appearsto be adapting well to regulatory and market developments, including both increaseddemand and potentially reduced reuse due to central clearing of OTC derivatives.Again, however, the distribution of eligible collateral holdings is important, and somemarket participants may still face liquidity constraints.

The RBA’s collateral eligibility criteria for its market operations permit a broaderrange of assets to be used than AGS and semi-government securities. If market par-ticipants faced material collateral constraints, greater use could potentially be madeof RBA-eligible nongovernment securities. There is some evidence that this has beenoccurring. More broadly, while we acknowledge that many other types of assets –including foreign assets or less liquid assets, such as corporate bonds – can be usedas collateral in some transactions, for many uses only high-quality Australian dollarassets are eligible. The fact that the survey found that use of GC 2 is quite limitedunderscores this point. It is for this reason that this paper focuses only on high-qualityAustralian dollar assets.

Finally, the paper highlights the potentially important role of collateral reuse inboosting effective supply, particularly in markets, such as Australia, that have rela-tively low active supply. Seeking to address concerns around the financial stabilityrisks associated with collateral re-hypothecation and reuse by placing tight restrictions





46 B. Cheung et al

on such activity could therefore be counterproductive. This lends support to the find-ings of recent work by the FSB, which instead focused on greater transparency ofsecurities lending and repo activity, and in particular collateral reuse (Financial Stabil-ity Board 2016). Data such as that collected by the RBA to inform this paper could bea useful addition to policymakers’ information set in this area. Indeed, as noted, sinceit is difficult to draw definitive conclusions from a limited number of observations,regular data collection to develop a time series would be informative as the marketadapts to a new regulatory and operating environment.



ACKNOWLEDGEMENTS

The authors would like to acknowledge the valuable input to this paper of a number ofcolleagues, and in particular Matthew Boge and Nicholas Garvin. This paper updatesan earlier version published in the Reserve Bank of Australia Bulletin (see Cheunget al 2014).

REFERENCES

Australian Prudential Regulatory Authority, Australian Securities and Investments Commis-sion and Reserve Bank of Australia (2014). Report on the Australian OTC derivativesmarket. Report, April, APRA/ASIC/RBA.

Australian Securities and Investments Commission (2015). ASIC derivative transactionrules (clearing). Report, ASIC.

Australian Securities and Investments Commission and Reserve Bank of Australia (2013).Implementing the CPSS–IOSCO principles for financial market infrastructures in Aus-tralia. Report, February, ASIC/RBA.

Basel Committee on Banking Supervision and the International Organization of SecuritiesCommissions (2015). Margin requirements for non-centrally cleared derivatives. Report,Bank for International Settlements, Basel.

Capel, J., and Levels, A. (2012). Is collateral becoming scarce? Evidence for the euro area.The Journal of Financial Market Infrastructures 1(1), 29–53 (http://doi.org/bj7r).

Cheung, B., Manning, M., and Moore, A. (2014). The effective supply of collateral inAustralia. RBA Bulletin, September, pp. 53–66, Reserve Bank of Australia.

Committee on the Global Financial System (2013). Asset encumbrance, financial reformand the demand for collateral assets. CGFS Papers 49, Bank for International Settle-ments, Basel.

Committee on Payment and Settlement Systems and Technical Committee of the Inter-national Organization of Securities Commissions (2012). Principles for financial marketinfrastructures. Report, Bank for International Settlements, Basel.






Cruz Lopez, J., Mendes, R., and Vikstedt, H. (2013). The market for collateral: the poten-tial impact of financial regulation. Financial Stability Review, June, pp. 45–53, Bank ofCanada.

Debelle, G. (2011).The committed liquidity facility. Speech, APRA Basel III ImplementationWorkshop, Sydney, November 23.

Debelle, G. (2015). Some effects of the new liquidity regime. Speech, 28th AustralasianFinance and Banking Conference, Sydney, December 16.

Duffie, D., and Zhu, H. (2011). Does a central clearing counterparty reduce counterpartyrisk? Review of Asset Pricing Studies 1(1), 74–95 (http://doi.org/fdvj2x).

Duffie, D., Scheicher, M., and Vuillemey, G. (2015).Central clearing and collateral demand?Journal of Financial Economics 116(2), 237–256 (http://doi.org/bj7s).

European Central Bank (2013). Collateral eligibility requirements: a comparative studyacross specific frameworks. Report, July, ECB.

European Central Bank (2014). Euro repo market: improvements for collateral and liquiditymanagement. Report, July, ECB.

Fender, I., and Lewrick, U. (2013). Mind the gap? Sources and implications of supply–demand imbalances in collateral asset markets. BIS Quarterly Review, September, Bankfor International Settlements.

Financial Stability Board (2013). Policy framework for strengthening oversight and regula-tion of shadow banking. Final Policy Framework, August 29, FSB.

Financial Stability Board (2015). OTC derivatives market reforms: tenth progress report.Report, November, FSB.

Financial Stability Board (2016).Transforming shadow banking into resilient market-basedfinance: possible measures of collateral re-use. Report, February 23, FSB.

Fuhrer, L. M., Guggenheim, B., and Schumacher, S. (2015). Re-use of collateral in the repomarket. Working Paper 2/2015, Swiss National Bank.

Heath, A., and Manning, M. (2012). Financial regulation and Australian dollar liquid assets.RBA Bulletin, September, pp. 43–52, Reserve Bank of Australia.

Heath, A., Kelly, G., and Manning, M. (2013). OTC derivatives reform: netting and networks.In Liquidity and Funding Markets, Heath, A., Lilley, M., and Manning, M. (eds), pp. 33–73.Proceedings of a Conference, Reserve Bank of Australia, Sydney.

Heller, D., and Vause, N. (2012). Collateral requirements for mandatory central clearing ofover-the-counter derivatives. Working Paper 373, Bank for International Settlements.

International Swaps and Derivatives Association (2012). BCBS–IOSCO proposal on mar-gin requirements for non-centrally-cleared derivatives. Letter to BCBS and IOSCO,December 12, ISDA.

International Swaps and Derivatives Association (2014). Interest rate derivatives: aprogress report on clearing and compression. Report, February, ISDA.

International Swaps and Derivatives Association (2015). Margin survey: 2015. MarketResearch Report, August 11, ISDA.

Markit (2013). Securities finance review: Q1 2013. Research Report, May 10, Markit.

Mersch, Y. (2014). Towards a new collateral landscape. Speech, Second Joint CentralBank Seminar on Collateral and Liquidity, Hamburg, September 17, ECB/DeutscheBundesbank.





48 B. Cheung et al

Sidanius, C., and Zikes, F. (2012). OTC derivatives reform and collateral demand impact.Financial Stability Paper 18, Bank of England.

Singh, M. (2013). The changing collateral space. IMF Working Paper WP/13/25, Interna-tional Monetary Fund.

Wakeling, D., and Wilson, I. (2010).The repo market in Australia. RBA Bulletin, December,pp. 27–35, Reserve Bank of Australia.





We also offer daily, weekly and live news updates

Visit Risk.net/alerts now and select your preferences

Asset Management

Commodities

Derivatives

Regulation

Risk Management

CHOOSE YOUR PREFERENCES

Get the information you needstraight to your inbox

RNET16-AD156x234-NEWSLETTER.indd 1 21/03/2016 09:27





Research Paper

Mobilization of collateral in Germany as areflection of monetary policy and financialmarket developments

Alexander Müller, Jan Paulick, Jan Fichtner andHubert Wittenmayer

Deutsche Bundesbank, Postfach 10 06 02, 60006 Frankfurt am Main, Germany;emails: [email protected], [email protected],[email protected], [email protected]

(Received May 13, 2016; revised July 4, 2016; accepted July 12, 2016)

ABSTRACT

Participation in Eurosystem credit operations requires that credit institutions post col-lateral. Therefore, the development of deposited collateral reflects changes in finan-cial markets and monetary policy. This paper describes and analyzes – for the periodFebruary 2008 to March 2016 – developments in the market value of marketableassets submitted as collateral in Germany and the Eurosystem against the backdropof the financial market crisis. The development is characterized by an initial strongincrease at the onset of the crisis and a decrease after 2010 because of lower fundingrequirements. The posted collateral followed the course of the funding requirements,which initially rose sharply in the wake of the financial crisis. Due to high liquidityinflows, which were reflected in the increasing TARGET2 claims of the Bundes-bank, the refinancing needs and posted collateral decreased after 2010. However, theposted collateral relative to refinancing operations remained remarkably high. Credit

Corresponding author: J. Paulick Print ISSN 2049-5404 j Online ISSN 2049-5412Copyright © 2016 Incisive Risk Information (IP) Limited

49




50 A. Müller et al

institutions can use different submission channels for collateral. We identify signifi-cant shifts among these channels. While these shifts are partly due to more technicalaspects, they may also stem from a “home bias” and portfolio reallocations.

Keywords: collateral; mobilization channels; Eurosystem; monetary policy; financial system.

1 INTRODUCTION

In order to participate in Eurosystem credit operations, credit institutions must pro-vide the Eurosystem with collateral. This paper describes and analyzes developmentsin the market values of marketable assets posted as collateral in Germany and theEurosystem against the backdrop of the financial market crisis for the period fromearly 2008 to 2016. These developments initially saw refinancing requirements risesharply before falling again after 2010, and they had a considerable impact on col-lateral holdings. We identify shifts between mobilization channels, at least some ofwhich may be due to an increase in home bias and portfolio shifts alongside technicalaspects.

In times of very low interest rates, the characteristics of collateral frameworks,the channels of collateral mobilization and the modes of collateral utilization allincrease in importance significantly (see Belke 2015; Nyborg 2015). This paper isa contribution to a growing body of research on aspects of collateral. In particular,we focus on the question of how monetary policy is transmitted by and reflected inthe mobilization of collateral over time, both concerning the market value of pledgedcollateral and the utilization of different submission channels. This paper proceedsas follows. Section 2 analyzes the development of pledged collateral regarding themarket value of deposited securities. In Section 3, we present findings on how theutilization of different mobilization channels has changed from 2008 to 2016. Finally,Section 4 concludes.

2 DEVELOPMENTS IN THE MARKET VALUES OF SUBMITTEDCOLLATERAL SINCE 2008

Counterparties’ collateral holdings at the Bundesbank were increased at the outbreakof the financial crisis, which in turn increased their potential for acquiring liquiditythrough Eurosystem monetary policy operations. When Lehman Brothers collapsed inSeptember 2008, the total volume of assets held for refinancing purposes rose sharply.The total volume of marketable assets deposited with the Bundesbank climbed from€600 billion in January 2008 to €770 billion in September 2009, a high level at whichit remained until the start of 2010.





Mobilization of collateral in Germany 51

FIGURE 1 Developments in refinancing operations and the market value of marketableassets submitted to the Bundesbank as collateral.

10

15

20

25

30

35

40

45900

800

700

600

500

400

300

200

100

0 0

Bill

ions

(€

)

5

%

Jan2008

July2008

Jan2009

July2009

Jan2010

July2010

Jan2011

July2011

Jan2012

July2012

Jan2013

July2013

Jan2014

July2014

Jan2015

July2015

Ratio of refinancing to submitted collateral (right-hand scale)

Refinancing operationsMarket value of submitted marketable assets

The refinancing operations presented here include main refinancing operations and longer-term refinancingoperations. Data shows monthly averages.

Immediately after the collapse of Lehman Brothers, conditions on the money mar-ket deteriorated dramatically. Liquidity and solvency problems at individual creditinstitutions led to a major loss of trust between banks. General levels of uncertaintyin the financial sector escalated, as did market participants’ concerns regarding suffi-cient liquidity. This resulted in the segmentation of the interbank money market intofinancial institutions with liquidity surpluses and financial institutions with liquiditydeficits, which could no longer be offset against each other.1

At the end of 2008, the Eurosystem took various measures to stabilize financial mar-kets, which had an impact on the submission of collateral. First of all, the Eurosystem’smain refinancing operations were changed from variable rate tenders to fixed rate ten-ders with full allotment. This increased the refinancing volume, which, in turn, raisedthe amount of collateral required. Moreover, the list of eligible assets was expanded

1 For an overview of the effects of the collapse of Lehman Brothers on the interbank market, seeDeutsche Bundesbank (2009, 2015a) and European Central Bank (2009).





52 A. Müller et al

by lowering the minimum credit threshold for marketable and nonmarketable assetsfrom A to BBB (except for asset-backed securities), effective from October 22, 2008(European Central Bank 2008a, 2008b). This measure counteracted a potential short-age of high-quality assets. Figure 1 clearly shows an increase in the mobilization ofcollateral in connection with the higher refinancing volume at the end of 2008 and in2009, after which the further posting of collateral mirrors the rise in funding require-ments. This is particularly apparent from the brief rise in the ratio of refinancing tosubmitted collateral in October 2008. Since the middle of 2010, there has been anoverall decline in the ratio of refinancing operations to submitted collateral, whichcurrently stands at around 10%. The market value of the mobilized collateral in rela-tion to the volume of refinancing operations is therefore fairly high, which may bedue in part to a lagged adjustment of the submission of collateral to the refinancingrequirements.

In addition, it should be noted that banks generally pledge collateral with a highermarket value to central banks than is really needed. This excess collateral can then beused for intraday TARGET2 operations or for potential margin calls. This is likely oneimportant reason why the market value of pledged collateral is always considerablyhigher than bank borrowing in the Eurosystem. Further, private banks frequently postcomparably illiquid and low-quality collateral to the domestic central bank. This isrational for banks, because such risky securities may either not be accepted by otherprivate banks at all or, if they are accepted, they may only be used under unfavorableconditions (see Belke 2015; Nyborg 2015).

The increase in marketable assets in 2008 shown here was accompanied by anincrease in the mobilization of nonmarketable assets, whose value almost doubledfrom around €40 billion in mid-2008 to €78 billion in mid-2009. Since then, thisfigure has remained roughly stable. As a result, only changes in the submission ofmarketable assets are analyzed here. For simplicity, the value of collateral providedis calculated based on its market value and not on its (lower) lending value. This doesnot significantly distort the trends, as in the period under observation it only resultedin changes in the haircuts to be applied to the market values in the case of specificsubsections of asset classes in the eligible collateral pool.

In the second quarter of 2010, the international financial and economic crisis turnedinto a combined and mutually reinforcing sovereign debt and banking crisis in Europe.There was a sharp decline in longer-term refinancing operations in mid-2010, owingto the expiry of the first one-year tender. In December 2011 and February 2012,for the first time, the Eurosystem provided commercial banks with liquidity throughlonger-term refinancing operations (LTROs), with maturities of up to thirty-six months(European Central Bank 2011). However, this measure was not accompanied by a risein the collateral deposited by the Bundesbank’s counterparties. On the contrary, thecollateral stock almost halved between September 2009 and the beginning of 2014.






FIGURE 2 Scatter plot of refinancing operations and posted marketable assets (monthlyaverages).

0

Sub

mitt

ed m

arke

tabl

e as

sets

(mar

ket v

alue

in €

bill

ions

)

Refinancing operations (in € billions)

100 200 3000

100

200

300

400

500

600

700

800

900

Since 2012, however, the value of assets posted as collateral has been becoming morestable, which reflects the overall decline in the ratio of refinancing operations to postedcollateral. Testing for a structural break in the development of refinancing operationsand collateral using a supremum Wald test, we found a statistically significant breakin December 2011.

Figure 2 shows the strong correlation (0.93) between refinancing operations andcollateral submitted. The results of Granger tests suggest that there is a time lag in theadjustment of the collateral holdings to the more volatile refinancing requirements,but not the other way around (see Box 1).

German banks’ample liquidity is in part the result of liquidity inflows, most notablyfrom the peripheral countries, which are reflected in the Bundesbank’s elevated TAR-GET2 claims (European Central Bank 2015a). This means that German commer-cial banks needed to rely less strongly on refinancing loans to obtain central bankmoney because they received central bank money through transfers from the euroarea (Deutsche Bundesbank 2011).

The resulting improvement in the banks’liquidity positions means a reduction in theamount of collateral they need to hold for refinancing operations. We can thus identifya contrasting development in the volume of pledged collateral and the Bundesbank’spositive TARGET2 balance between 2012 and 2014 (see Figure 3).

The covered bond purchase program (CBPP), the asset-backed securities purchaseprogram (ABSPP) and the secondary markets public sector purchase program (PSPP)adopted by the European Central Bank Governing Council are also likely to have





54 A. Müller et al

BOX 1 The relationship between refinancing operations and posting of collateral.

The Granger causality test determines for two stationary time series whether, after a timelag, one time series is useful in forecasting the other. However, in contrast with what itsname suggests, the test cannot determine causality (see Granger 1969; Lütkepohl 2005).For the relationship between refinancing operations and the posting of collateral, Grangercausality tests provide evidence that the collateral stock can be (partly) forecast using therefinancing volume, but not vice versa. This implies that collateral submission is adjustedaccording to refinancing needs. The alternative hypothesis would imply that collateral isadjusted in anticipation of refinancing operations; however, this hypothesis of forward-looking collateral submission is rejected.

Results of vector autoregressive (VAR) regressions with one lag are presented inTable 2. Interms of the selection of the number of lags, we get conflicting results when using differentinformation criteria. We focus on the results for one time lag, but the results are similarwhen including more lags. To achieve a stationary time series, we use first differences. Asa robustness check, we find the results for levels to be similar.

The residuals are not normally distributed, due to an outlier in October 2008. This stemsfrom the fact that in October 2008 eligible securities were expanded, and there was a switchfrom rate tenders to full allotment.The date can therefore be regarded as an external shockor a structural break in the data.As further robustness checks, we test for Granger causalityexcluding October 2008, for the periods before and after October 2008 and for the periodsbefore and after December 2011, for which a structural break was identified from the data(see Table 3). We find the dynamics to be unaffected, with significance slightly decreasingwith more limited sample sizes.The results always remain significant at a 5% level and arenever significant at conventional levels for the alternative null hypothesis.

The results show that refinancing operations have predictive power for submitted collateralin the next period, with a positive coefficient. Arguably, the reason for this is that collateralis adjusted upward following periods with higher refinancing and downward following peri-ods with less refinancing operations. This indicates that the decline in posted collateral islargely the result of the decrease in liquidity requirements over time, but not vice versa.Therefore, the potential forward-looking collateral submission for future refinancing needsis discarded.

TABLE 1 Tests of hypotheses for Granger-causality (one lag).

Null hypotheses F -statistic Probability Result

H0: refinancing operations (RO) donot Granger-cause submittedmarketable assets

12.70 0.001 Reject H0

H0: submitted marketable assets(SMAs) do not Granger-causerefinancing operations

0.82 0.367 Accept H0






TABLE 2 VAR results (first difference on monthly averages).

(1) (2) (3) (4)Variables �RO �SMA �RO �SMA

�ROt�1 0.001 0.241*** — 0.271***(0.010) (3.564) — (4.100)

�SMAt�1 �0.137 0.163* �0.136 —(�0.907) (1.698) (�0.941) —

Constant �1.623eC09 �1.356e+09 �1.623eC09 �1.669eC09(�0.741) (�0.972) (�0.746) (�1.194)

Observations 97 97 97 97Adjusted R-squared �0.012 0.158 �0.001 0.141Probability > F 0.647 0.000 0.349 0.000

t -statistics in parentheses: ***p < 0.01, **p < 0.05, *p < 0.1.

contributed to lower funding requirements among counterparties (European CentralBank 2015b).

With regard to the scale of the PSPP, ABSPP and CBPP3, a monthly purchase vol-ume of €60 billion was originally foreseen; in April 2016 this was increased to €80billion per month. In addition, in June 2016, the Eurosystem central banks have begunthe corporate sector purchase program (CSPP), buying bonds issued by nonbank cor-porations established in the euro area that have a sufficient credit rating (investmentgrade) (European Central Bank 2016a). The CSPP – amounting to €6.4 billion forJune 2016 – will contribute to the asset purchase program’s average monthly purchasevolume of €80 billion (European Central Bank 2016b). The initially adopted purchas-ing window from March 2015 (or from October and November 2014 for CBPP3 andABSPP) until 2016 has been extended to at least March 2017 (European Central Bank2016c). The extensive provision of liquidity via these securities programs could meanthat counterparties avail themselves of an even smaller refinancing volume at the Bun-desbank, and thus pledge less eligible collateral (Deutsche Bundesbank 2015b). Atthe moment, however, there are no signs of a decline, which could potentially beattributable to the lagged adjustment in the submission of collateral. An importantdevelopment to observe will be the future availability of high-quality collateral, suchas German government bonds, whose yields have turned negative even for the ten-yearbonds in mid-June 2016.

A comparison of the eligible collateral submitted to the central bank in Ger-many over time with that submitted in the rest of the Eurosystem (excluding Ger-many) reveals that after the onset of financial market turmoil in August 2007 andthe Lehman Brothers insolvency in September 2008, the volume of collateral heldinitially increased sharply in both cases (see Figure 4).





56 A. Müller et al

TAB

LE

3R

obus

tnes

sch

ecks

for

diffe

rent

time

perio

ds.

Nu

llhy

po

thes

esR

obu

stn

ess

chec

kO

bse

rvat

ion

sF

-sta

tist

icP

rob

abili

ty

H0:

RO

does

notG

rang

er-c

ause

SM

AO

ctob

er20

08ex

clud

ed96

7.77

0.00

7H

0:S

MA

does

notG

rang

er-c

ause

RO

Oct

ober

2008

excl

uded

960.

640.

424

H0:

RO

does

notG

rang

er-c

ause

SM

AU

pto

Oct

ober

2008

811

.15

0.02

1H

0:S

MA

does

notG

rang

er-c

ause

RO

Up

toO

ctob

er20

088

2.42

0.18

1H

0:R

Odo

esno

tGra

nger

-cau

seS

MA

Afte

rO

ctob

er20

0889

8.60

0.00

4H

0:S

MA

does

notG

rang

er-c

ause

RO

Afte

rO

ctob

er20

0889

0.45

0.50

2H

0:R

Odo

esno

tGra

nger

-cau

seS

MA

Up

toD

ecem

ber

2011

466.

170.

017

H0:

SM

Ado

esno

tGra

nger

-cau

seR

OU

pto

Dec

embe

r20

1146

0.22

0.64

2H

0:R

Odo

esno

tGra

nger

-cau

seS

MA

Afte

rD

ecem

ber

2011

515.

790.

020

H0:

SM

Ado

esno

tGra

nger

-cau

seR

OA

fter

Dec

embe

r20

1151

2.23

0.14

2






FIGURE 3 Market value of marketable assets deposited with the Bundesbank andTARGET2 balance.

Market value of submitted marketable assetsTARGET2 claims

800

700

600

500

400

300

200

100

0

Bill

ions

(€

)

Jan2008

Jan2009

Jan2010

Jan2011

Jan2012

Jan2013

Jan2014

Jan2015

Jan2016

FIGURE 4 Comparison over time of eligible collateral deposited with the Bundesbankand the Eurosystem central banks (excluding the Bundesbank).

02008 2010 2012

Q12012Q3

2013Q1

2013Q3

2014Q1

2014Q3

2015Q1

2016Q1

2015Q3

500

1000

1500

2000

2500

Bill

ions

(€

)

Bundesbank European central banks (excluding Bundesbank)

Eurosystem data from 2008 to 2011 was only available on an annual basis. Source: European Central Bank (www.ecb.europa.eu/paym/coll/charts/html/index.en.html) and own calculations.





58 A. Müller et al

FIGURE 5 Distribution of the volume of marketable assets among the various mobilizationchannels (as at end-March 2016).

Domestic XemacCCBM Third-party custody

BOX 2 Mobilization channels of marketable assets.

There are four channels via which marketable assets may be submitted in the form of apledge to the Bundesbank (see also Table 3).

(1) Domestic securities pledge (domestic): eligible assets held at Clearstream Bank-ing Frankfurt (CBF) are individually transferred to the Deutsche Bundesbank’s safecustody account held at CBF, or transferred via eligible links to the DeutscheBundesbank’s safe custody account at CBF or Clearstream Banking Luxembourg(CBL).

(2) Securities deposited at a domestic custodian bank may be transferred individuallyto the Deutsche Bundesbank within the framework of third-party custody.

(3) The eligible securities deposited at CBF or CBL may also be made available in favorof the Deutsche Bundesbank as a global amount via Xemac (triparty system), CBF’scollateral management system. The cross-border mobilization of collateral is alsopossible via other triparty services (CmaX from Clearstream Banking Luxembourgand AutoSelect from Euroclear).

(4) Eligible securities deposited in other Eurosystem member states may be used ona cross-border basis under the CCBM. The national central banks maintain safecustody accounts with each other for this purpose.

However, in the ensuing period, a countervailing trend set in. The fall that can beobserved in Germany as of 2009 reflects already improved liquidity conditions forGerman institutions. In the rest of the Eurosystem, on the other hand, the mobilizationof collateral continued to rise after 2009 until it reached an initial peak in 2010, beforefalling slightly in 2011. Following this, there was again a strong rise in the value of






TABLE 4 Mobilization channels of eligible assets to the Deutsche Bundesbank.

Method Custodian bank Transmission

Domestic CBF Individually to the DeutscheBundesbank’s safe custodyaccount at CBF or CBL

Third-party safecustody

Domestic custodian bank Individually to the DeutscheBundesbank’s safe custodyaccount

Triparty systems CBF, CBL, Euroclear FranceS.A./Euro-clear Bank S.A.(ECL)

Provision of global amounts(securities claim amounts)via Xemac, CmaX orAutoselect

CCBM Foreign central bank Individually to the foreigncentral bank’s safe custodyaccount in favor of theDeutsche Bundesbank

pledged collateral, above all to enable participation in the two three-year LTROs inDecember 2011 and February 2012. Demand amounted to approximately €500 billionon both of these allotment dates.

In absolute figures, the total amount of collateral deposited by the counterparties inthe rest of the Eurosystem climbed from €950 billion in 2008 to around €2.1 trillionin the third quarter of 2012. As of the first quarter of 2013, holdings of collateral fellcontinuously until the start of 2015, at which time they stood at €1.4 trillion.

3 CHANGES OVER TIME IN MOBILIZATION CHANNELS

At present at the Bundesbank, 38% of pledged marketable collateral is submitted viaXemac, 27% is submitted via the domestic channel, 31% is submitted via the corre-spondent central banking model (CCBM) and 4% is submitted via third-party custody(for more information on these mobilization channels, see Box 2 and Figure 5).

All submission routes recorded increased directly following the outbreak of thefinancial crisis and the introduction of monetary policy measures and reached theirpeak – a total of €770 billion – in September 2009. Subsequently, structural shiftsoccurred to the mobilization channels used for pledging collateral (see Figure 6).

In particular, holdings with Xemac rose gradually, while a decline in domestic andCCBM submissions was observed. The volumes of marketable assets via each mobi-lization channel are shown individually in Figure 7. A shift between the individualchannels is clearly identifiable.





60 A. Müller et al

FIGURE 6 Volume of marketable assets by mobilization channel (aggregated).

0

100

200

300

400

500

600

700

800

Bill

ions

(€

)

Jan2008

Dec2008

Nov2009

Oct2010

Sep2011

Aug2012

Jul2013

Jun2014

May2015

Domestic Xemac CCBM Third-party custody

(a)(b) (c) (d) (e)(f) (g)

(i) (ii)

Data for third-party safe custody is only available as of November 2008. While triparty cross-border systems (seeBox 2) have been live since September 2014, they are not currently being used by counterparties of the Bundesbankto submit collateral to the bank and are therefore not yet included in the chart. (a) September 15, 2008: Lehmaninsolvency. (b) October 15, 2008: expansion of eligible securities; switch from variable rate tender to fixed rate tender(full allotment). (c) June 25, 2009: first twelve-month tender. (d) December 17, 2009: last twelve-month tender.(e) December 22, 2011: first three-year tender. (f) February 29, 2012: second three-year tender. (g) September 29,2014: cross-border triparty services. (i) Highest level in September 2009: €770 billion. (ii) Lowest level in March2016: €380 billion.

Particularly noteworthy is the growing importance of the Xemac system, whichhas evolved from a relatively insignificant submission channel in 2008 to the largestone in 2016.

With respect to the increasing use of the Xemac mobilization channel over time,technical features, such as the ease of use of securities holdings for several purposesbeyond monetary policy operations, might play a role.

In September 2008, around half of all securities submitted were posted via theCCBM submission channel, which clearly dominated until 2009 before subsequentlyrecording the lowest growth in both relative and absolute terms. The submission ofcollateral via the CCBM channel had dropped even shortly after the crisis began,before the total amount of deposited collateral had reached its peak. The lower useof this mobilization channel by counterparties of the Bundesbank could be due to apreference for domestic securities (home bias) and a redistribution of collateral stock.It is possible that the Deutsche Bundesbank’s counterparties reduced their securities






FIGURE 7 Volume of marketable assets by mobilization channel (not aggregated).

0

50

100

150

200

250

300

350

Bill

ions

(€

)

Jan2008

Dec2008

Nov2009

Oct2010

Sep2011

Aug2012

Jul2013

Jun2014

May2015

Domestic Xemac CCBM Third-party custody

(a)(b) (c) (d) (e)(f) (g)

(i) (ii)

Financial crisisspills over into

a sovereigndebt crisis

The third-party custody of collateral was taken into account in the statistics for the domestic submission route untilNovember 2008. (a) September 15, 2008: Lehman insolvency. (b) October 15, 2008: expansion of eligible securities;switch from variable rate tender to fixed rate tender (full allotment). (c) June 25, 2009: first twelve-month tender.(d) December 17, 2009: last twelve-month tender. (e) December 22, 2011: first three-year tender. (f) February 29,2012:second three-year tender. (g) September 29, 2014:cross-border triparty services. (i) Highest level in September2009: €770 billion. (ii) Lowest level in March 2016: €380 billion.

holdings due to the credit ratings of the government debt instruments of affected euroarea periphery countries being downgraded and their prices falling. The use of CCBMreached its lowest point to date in summer 2014, when it slumped to €98 billion, lessthan one-third of the value recorded in spring 2008.

In November 2009, third-party custody plummeted from around €55 billion to€25 billion. This abrupt drop is attributable to the discontinuation of third-partycustody services by one third-party custodian. By March 2016, the value of securitiesheld in third-party custody had been reduced once again by almost half and stood at€13 billion.

4 CONCLUSION

The increase in the overall stock of collateral held by the Deutsche Bundesbank isclearly indicative of the changes taking place in financial markets. Collateral stockdevelopments mirror funding requirements to a certain extent, with the latter initially





62 A. Müller et al

rising sharply on account of the financial crisis. Using Granger causality tests, we findthat refinancing operations have predictive power for future collateral mobilization,but not vice versa. Due to high liquidity inflows, which were reflected in the Bun-desbank’s escalating TARGET2 claims, funding requirements – and hence collateralstock – fell. The extensive provision of liquidity via the securities purchase programscould lead to a further reduction in overall collateral holdings in the coming months.Shifts between mobilization channels were largely caused by technical aspects. How-ever, the decline in CCBM could also be attributable to counterparties shifting theirportfolios.


This paper represents the judgments and views of the authors and does not necessarilyreflect the opinion of the Deutsche Bundesbank.

REFERENCES

Belke, A. (2015). Eurosystem collateral policy and framework: post-Lehman time as a newcollateral space. Intereconomics 50(2), 82–90 (http://doi.org/bm7k).

Deutsche Bundesbank (2009). Financial stability review 2009. Report, Deutsche Bundes-bank.

Deutsche Bundesbank (2011). Annual report 2011. Report, Deutsche Bundesbank.Deutsche Bundesbank (2015a). Germany in the financial and economic crisis. Monthly

Report, October, Deutsche Bundesbank.Deutsche Bundesbank (2015b).Marketable financial instruments of banks and their role as

collateral in the Eurosystem. Monthly Report, June, pp. 31–44, Deutsche Bundesbank.European Central Bank (2008a). Measures to further expand the collateral framework and

enhance the provision of liquidity. Press Release, October 15.European Central Bank (2008b). Technical specifications for the temporary expansion of

the collateral framework. Press Release, October 17.European Central Bank (2009).The ECB’s response to the financial crisis.Monthly Bulletin,

October, pp. 59–74.European Central Bank (2011). ECB announces measures to support bank lending and

money market activity. Press Release, December 8.European Central Bank (2015a). Publication of TARGET balances. Economic Bulletin 6,

42–44.European Central Bank (2015b). Introductory statement. Press Conference Speech,

December 3.European Central Bank (2016a). European Central Bank announces details of the corpo-

rate sector purchase programme (CSPP). Press Release, April 21.European Central Bank (2016b). Asset purchase programmes. URL: www.ecb.europa.eu/

mopo/implement/omt/html/index.en.html.European Central Bank (2016c). Introductory statement.Press Conference Speech, March

10.






Granger, C. W. J. (1969). Investigating causal relations by econometric models and cross-spectral methods. Econometrica 37(3), 424–438.

Lütkepohl, H. (2005). New Introduction to Multiple Time Series Analysis. Springer (http://doi.org/dzdq3t).

Nyborg, K. G. (2015). Central bank collateral frameworks. Discussion Paper 10663, Centrefor Economic Policy Research, London.






Forum Paper

Collateral flows and balance sheet(s) space

Manmohan Singh

International Monetary Fund, 1900 Pennsylvania Avenue NW, Washington, DC 20431, USA;email: [email protected]

(Received May 17, 2016; revised July 15, 2016; accepted July 15, 2016)

ABSTRACT

Collateral does not flow in a vacuum; it needs balance sheets to move within thefinancial system. The new regulations constrain private sector bank balance sheetsand thus impede market plumbing. This paper looks at securities-lending, derivativesand prime-brokerage markets as suppliers of collateral (as much has been written onthe repo market). Going forward, the choice of balance sheet(s), private or public,should be driven by market forces and not by the ad hoc allocation of central banks.Otherwise, this may be suboptimal for monetary policy transmission.

Keywords: collateral flows; securities lending; prime-brokerage; derivatives; balance sheets;monetary policy transmission.

1 INTRODUCTION

Collateral flows lie at the heart of any proper understanding of market liquidity and,hence, financial stability. No other market is so critical to the functioning of thefinancial system and yet so poorly understood. In addition, though, as policy makersbegin to acknowledge the inadequacies of traditional theories of money and lending,collateral flows are increasingly being recognized as a driver of credit creation that isjust as important as money itself. Despite this, a true appreciation of the importance

Print ISSN 2049-5404 j Online ISSN 2049-5412Copyright © 2016 Incisive Risk Information (IP) Limited

65




66 M. Singh

of collateral flows is hampered by the inadequacy of the way in which they areaccounted for.

For overall financial lubrication, the financial system requires collateral or moneyfor intraday debits and credits. The cross-border financial markets traditionally use“cash or cash-equivalent” collateral (ie, money or highly liquid fungible securities)in lieu of cash to settle accounts. Financial collateral does not have to be highly rated(AAA/AA): as long as the securities (which can be either debt or equity) are liquid,mark-to-market and part of a legal cross-border master agreement, they can be usedas “cash equivalent”. In this way, collateral underpins a wide range of secured fund-ing and hedging (primarily with over-the-counter (OTC) derivatives) transactions.Increasingly, collateral has a regulatory value in addition to being cash-equivalent.Such financial collateral has not yet been quantified by regulators and is not (yet)part of official sector statistics; however, it is a key component of financial plumbing(Baklanova et al 2016).

1.1 The discomfort with “collateral chains”

The term “pledged for reuse” means that the collateral taker has the right to reuse itin their own name. Its practical effect is economically equivalent to title transfer (ie, achange in ownership) and is essential to the financial lubrication that makes collateralakin to cash-equivalent. In the bilateral market, contracts that embrace repurchaseagreement (repo), securities-lending, OTC derivatives and customer margin loansmay involve title transfer. Under a title-transfer arrangement, the collateral providertransfers ownership of collateral to the collateral taker.

The latter acquires full title to the collateral received and, as its new owner, iscompletely free to utilize it. In return, the parties agree that, once the collateral providerhas discharged its financial obligation to the collateral taker, the collateral taker willreturn equivalent collateral to the collateral provider. Note that the obligation is toreturn equivalent collateral: that is to say, securities of the same type and value terms,but not the original security. This point about equivalence is important. After thecollateral has changed hands via title transfer and been reused by the collateral taker,it would not be obligatory on the part of the collateral taker to return exactly the sameproperty initially received as collateral. A simplistic example is a physical twenty-dollar bill with serial number XYZ. If you provide that very bill as collateral to thecollateral recipient, it does not matter if they give you back a different twenty-dollarbill – any twenty-dollar bill will do.

Although the terms “rehypothecation” and “pledged collateral that can be reused”are often employed interchangeably, each has a specific and slightly different meaning.“Rehypothecation” means the use of financial collateral by a collateral taker as securityfor their own obligations to some third party (ie, onward pledging). Reuse is broader in





Collateral flows and balance sheet(s) space 67

scope, encompassing not only repledging but also any use of the collateral compatiblewith ownership of the property (such as selling or lending it to a third party). Not allpledged collateral can be reused in this way. Rights of reuse are thus inherent in title-transfer financial collateral arrangements, because ownership of the property actuallychanges, whereas under a pledge the collateral taker takes a security interest only in thepledged assets, and they will enjoy rights of rehypothecation only if reuse is expresslygranted in the pledge agreement.1 Market practice suggests that rehypothecation ofassets has historically been a cheaper way of financing the prime business than turningto the repo market, and some of the recent regulations are more beneficial to nettingfor prime brokerage (eg, equity long/short positions) than repo.

Within the United States, rehypothecation rights are strictly limited. Outside theUnited States (that is, outside New York-governed contracts), the prevalence of rehy-pothecation allows for a market clearing price for financial collateral (ie, the UnitedKingdom and continental Europe). Rights of reuse have a strong legal underpin-ning under the Financial Collateral Directive of the EU. The EU legal framework forfinancial collateral is flexible and can accommodate the preferences of prudent andrisk-averse clients and counterparties. Whether or not sophisticated market partici-pants strike bargains that offer them appropriate protection is a matter for them aloneto decide. In most cases, UK broker-dealers operate subject to contractually agreedreuse limits (see Appendix A).

Some policy makers, especially in the financial stability groups (eg, the Finan-cial Stability Board (FSB)), perceive “rehypothecation” to be systemically dangerous(because of the way it can drive leverage). However, ordinary banking is not funda-mentally different. In economic terms, the “reuse” or rehypothecation of a securityis identical to the money creation that takes place in commercial banking throughthe process of accepting deposits and making loans. So, why is it that a deposit at abank of US$100 dollars can be lent, but financial collateral that is mark-to-market atUS$100 dollars is restricted for reuse by policy makers? A bank such as Citibank hascapital; so does shadow banking via haircuts and overcollateralization whenever col-lateral is reused. Central banks are trying to rejuvenate the credit-creation engine viaquantitative easing (QE); so far, they are not having great success. Monetary policy

1 Under a pledged collateral agreement, the collateral taker, or the “pledgee”, does not have automaticrights of reuse or rehypothecation in the pledge agreement unless such rights of reuse are expresslygranted in the contract. The pledgee will not be able to seize or use that pledged collateral fortheir own purposes unless the “pledgor” defaults on their obligation to the pledgee, triggeringenforcement. However, in cases where a pledgor, or collateral provider, grants a pledgee rightsof rehypothecation over pledged collateral, and if the pledgee has exercised this right prior toinsolvency, the pledgor’s legal rights are as if they had transferred title in the property to the pledgee.The pledgor’s legal remedies against an insolvent pledgee are, in practice, extremely limited.





68 M. Singh

is ultra loose. Restricting collateral reuse is a tight monetary policy that seems to beat odds with the current policies of key monetary authorities. In fact, the money met-rics such as M0, M1 and M2 need to integrate the sizable pledged collateral metrics.Otherwise, fully understanding the financial plumbing (that accepts both money andpledged collateral as lubricants) will not be possible.

2 QUANTITATIVE EASING AND REGULATIONS

Expanded central bank balance sheets that silo sizable holdings of US treasuries, UKgilts, Japanese government bonds (JGBs), German bunds and other AAA eurozonecollateral have placed central bankers in the midst of market plumbing. It is now goingto be very difficult for them to walk away from that role. For example, had QE nothappened then deposits would have grown roughly in line with the economy’s growthand/or household wealth. However, in the United States, where QE has ended, datafrom June 2015 shows that deposits with the Federal Deposit Insurance Corporation(FDIC) have doubled in the top fifty US bank holding companies relative to June2008 levels. The eurozone and Japan are in the midst of their QE at present.

Given the near double digit return that global systemically important financialinstitutions (G-SIFIs) need for their shareholders, some deposits are being pushedout to the official sector balance sheet; otherwise, these deposits would be a dragfor the banks and result in lower returns for their shareholders. In other words, theexcess deposits (stemming from nonbank sales of collateral to the central banks) andforthcoming regulations, such as the leverage ratio, which effectively requires banksto hold capital against deposits, are too “costly” for banks; hence the reluctance bybanks to take these deposits on their balance sheet. A typical bank’s marginal returnon these sizable deposits is below their marginal return to their shareholders. Giventhe limited balance sheet space at the private sector banks, the demands for the officialsector (ie, central banks’) balance sheets will remain important unless regulations arefine-tuned to allow for more bank/nonbank intermediation.

The recent experience of the United States Federal Reserve (Fed) sheds some lighton the operational aspects that are relevant. For example, the taper tantrum of May2013 highlighted market volatility concerns; not surprisingly, the Fed’s liftoff decisionin December 2015 was associated with a large reverse-repo program (RRP), whichis a deft way of handling financial stability concerns stemming from losses and/orvolatility on longer-tenor US treasuries. Large foreign repo pools at the Fed (ie, thedeposits of foreign governments, central banks and international official institutions)and deposit accounts for central counterparties (CCPs) at central banks, etc, alsosuggest an expanded role for central bank balance sheets. However, financial plumb-ing, where money and collateral interface, is a role that has historically always beenassociated with private-sector market participants (ie, banks, nonbanks, custodians,






FIGURE 1 Bilateral pledged collateral flows via the key banks in this market.

0

200

400

600

800

1000

1200

0

200

400

600

800

1000

1200

1400

Bill

ions

US

$B

illio

ns U

S$

Bea

r S

tear

ns

2007 2008 2009 2010 2011 2012 2013 2014 2015

Lehm

an

Mor

gan

Sta

nley

Gol

dman

Sac

hs

Mer

rill/B

oA

JP M

orga

n

Citi

grou

p

(a)

(b)

Soc

. Gén

éral

e

BN

P P

arib

as

Deu

tsch

e B

ank

Cre

dit S

uiss

e

UB

S

Bar

clay

s

RB

S

HS

BC

Nom

ura

This includes collateral flows from repo, securities lending, prime brokerage and derivatives. (a) Pledged collateralreceived by US banks (2007–15). (b) Pledged collateral received by European banks and Nomura (2007–15).Source:annual report of banks (eg, 10 000 filings), hand-picked data by the author.

etc), not with central banks, whose mandate is about monetary policy. Market inter-est rates are effectively determined in the pledged collateral market, where banksand other financial institutions exchange collateral (such as bonds and equities) formoney.

In 2007, this global bilateral collateral market, where the plumbing takes place, wasUS$10 trillion in size; now it is well below US$6 trillion (see Figure 1 and also Box 1;note that the pledged collateral shown here is cross-border and not limited to reuse(unlike collateral within the triparty structure in the United States)). About half of the





70 M. Singh

pledged collateral comes from the hedge fund industry; the other source of pledgedcollateral is pensions, insurers, central banks, sovereign wealth funds (SWFs), etc(Singh 2011; European Systemic Risk Board 2014).

From Lehman’s last annual report:

At November 30, 2007, the fair value of securities received as collateral that werepermitted to sell or repledge was approximately [US]$798 billion.... The fair valueof securities received as collateral that were sold or repledged was approximately[US]$725 billion at November 30, 2007.

Pledged collateral from bilateral, securities-lending, prime brokerage, and OTCderivatives margin is hard to disentangle as it shows up bunched up in footnotes tobalance sheets. Collateral with title transfer is pooled at the central collateral desks atlarge banks (the top-tier G-SIFIs that have a global footprint). Major dealers active inthe collateral industry include Goldman Sachs, Morgan Stanley, JP Morgan, Bank ofAmerica/Merrill and Citibank in the United States. In Europe and elsewhere, importantcollateral dealers are Deutsche Bank, UBS, Barclays, Credit Suisse, Société Générale,BNP Paribas, HSBC, Royal Bank of Scotland and Nomura. This collateral with titletransfer (or a variant thereof) can come into the banks via reverse repo, securitiesborrowing or OTC derivatives margin posting, or the use of client assets under aprime-brokerage agreement. Thus, any collateral metric should capture the typicaldocumentation that underpins collateral use and reuse in contracts such as the globalmaster securities lending agreement (GMSLA), global master repurchase agreement(GMRA) and International Swaps and DerivativesAssociation (ISDA) agreement, etc.The documentation does not restrict collateral reuse to one jurisdiction (or region);hence, the collateral metric needs to be global.

Table 1 provides a succinct summary of the sources of collateral, the total volumereceived by the large banks and the resultant velocity. The velocity is not an exactmetric but gives an idea of the length of the collateral chains in that year. So, wecan infer that, on average, the collateral chains were longer in 2007 than in 2015.The intuition is that counterparty risk before the collapse of Lehman Brothers wasminimal. In the aftermath of Lehman’s demise, fewer trusted counterparties in themarket owing to elevated counterparty risk led to stranded liquidity pools; incompletemarkets; idle collateral; and shorter collateral chains, missed trades and deleveraging.At present, the collateral landscape has changed even more due to central banks’ QEpolicies, new regulations, etc. Collateral reuse (or velocity) is at an all-time low ofabout 1.8, compared with 3.0 before Lehman’s demise. This collateral reuse rate is acentral theme of this special edition of The Journal of Financial Market Infrastructuresand deserves more attention in policy circles (for example, the recent Jackson Holepapers straddled plumbing issues but were silent on collateral reuse rate).






BOX 1 The ten to fifteen banks at the core of global financial plumbing.

Let the financial system that includes banks, hedge funds, pension funds, insurers, SWFs,etc, be represented by A to Z. Only a handful (say XYZ) can move financial collateral acrossborders. XYZ also happen to be the large ten to fifteen banks discussed earlier. The rest ofthe financial system from A to W that demand and supply collateral need to connect with eachother via XYZ. Entry into this market is not prohibited but extremely expensive and difficult, aswe need a global footprint and global clients (and the acumen and sophistication to move andprice liquid securities very quickly, in seconds sometimes). For example, a Chilean pensionfund may want Indonesian bonds for six months, and W (for example, a hedge fund or asecurities lender in Hong Kong) may be holding these bonds and willing to rent out to A forsix months for a small fee. But W does not know there is demand from A. Only via XYZ canA connect to W. Since XYZ sit in the middle of the web, they have the ability to optimize inways that give them an advantage; the Indonesian bonds may come into their possessionbecause they have loaned W money, or because they have a derivative with W or through asecurity lending agreement.

Such securities that need to move cross-border under a “repo” or “security lending” or relatedtransaction need to be legally perfected (herein, legal perfection entails rules such as titletransfer and rehypothecation). Perfection is also possible under pledge, as documented inthe master securities lending agreement (MSLA). Similarly, for OTC derivative margins, thereis an ISDA master agreement. For prime-brokerage/hedge-fund collateral, there is a similarmaster agreement that resonates easily between XYZ.Thus, it is not easy for all real-economycollateral to be able to move across borders. This market for bilateral pledged collateral isthe only true market that prices at mark-to-market all liquid securities (bonds C equities).Given that collateral is in short supply (as reflected by repo rates), one of two things is likelyto happen.

(a) The velocity of collateral may come back: this is a task that only XYZ can handle inbulk if more good collateral is sourced through them. However, regulatory proposalssuch as leverage and liquidity ratio have resulted in balance-sheet constraints for XYZto do collateral transformation. So, the velocity or reuser rate is unlikely to come back(see Table 2).

(b) Central banks can make balance-sheet “space” in order to augment the balancesheets with XYZ, for example, the Fed’s reverse repo program since September 2013,which was augmented to almost US$2 trillion in December 2015. But this programdoes not release collateral to the market, as it uses the triparty structure; so, theFed’s counterparty gets ownership but not possession. This is one way to not let thecollateral “velocity” escape, which, in turn, would increase repo rates (and this mightcreate a wedge with the policy rate, so, conservatively, there is no leakage of durationto the market); thus, all maturing bonds bought under QE are reinvested. The Euro-pean Central Bank (ECB)-type of approach (that was seen during the EU crisis withsubsidized haircuts relative to market) may not be market based. More recently, inthe aftermath of the ECB’s QE since March 2015, its securities-lending program hasremained in its infancy. On the other hand, the Reserve Bank of Australia will not issuenew debt to meet collateral demand, but it will provide good collateral (or high-qualityliquid assets) at market price.





72 M. Singh

TABLE 1 Sources of pledged collateral, volume of market and velocity (2007; 2010–15).

Sources‚ …„ ƒ Volume ofHedge Securities secured Reuse rate

Year funds lending Total operations (or velocity)

2007 1.7 1.7 3.4 10.0 3.02010 1.3 1.1 2.4 5.8 2.42011 1.3 1.05 2.35 6.1 2.52012 1.8 1.0 2.8 6.0 2.22013 1.85 1.0 2.85 5.8 2.02014 1.9 1.1 3.0 5.8 1.92015 2.0 1.1 3.1 5.6 1.8

In trillions of US dollars; velocity in units. Sources: Risk Management Association (RMA); International MonetaryFund (IMF) Working Paper, “Velocity of pledged collateral” (Singh 2011).

3 EVIDENCE FROM SECURITIES-LENDING, DERIVATIVES ANDPRIME-BROKERAGE MARKETS

Much has been written about repo markets shrinking, but securities lending, deriva-tives and prime brokerage are also key avenues for collateral flows and reuse. Thus,the focus here is on these avenues, but repo (bilateral and triparty) is discussed suc-cinctly in Box 2. Collateral does not flow in a vacuum and thus needs balance sheetspace to move.

3.1 Securities lending

Although the large banks are unlikely to make room for the “high volume, low margin”securities lending business (due to leverage ratio), it is often assumed that the majorcustodians, such as Bank of NewYork (BNY) Mellon, Citibank, State Street, Euroclearand Clearstream, will have “balance sheet space” to move collateral around. Assetsheld by custodians are not part of their balance sheets; only principal positions areon these balance sheets. However, an indemnification requirement to clients requiresupfront capital provision, and this is not cost effective. Pre-Lehman, dealers wouldoblige the custodians that pushed out general collateral (eg, IBM or Merck equities)along with specials that the dealers really wanted (and still do). In this era, custodianswould set a general collateral (GC) to “specials” ratio as high as 5:1 or even 13:1;there was less balance sheet constraint. For almost a decade now, there has been notying of GC to specials.

The asset-management complex, which includes pensions, insurers and official sec-tor accounts such as SWFs and central banks, is a rich source of collateral deposits. Thesecurities they hold are continuously reinvested (via securities lending) to maximize






returns over their maturity tenor. In a repo, there is an outright sale of the securities,which is accompanied by the specific price and date at which the securities will bebought back. On the other hand, securities-lending transactions generally have no setend date and no set price, although the market for defined-term trades is growingsecurities.2 Borrowing is generally done with a specified purpose, and in many casesa legal purpose test is required. As such, securities-lending markets are utilized toborrow specific securities, whereas repo markets are generally non-security specific.In 2007, securities-lending volumes were US$1.7 trillion. In recent years, despitecollateral constraints, the volumes are flat at around US$1 trillion, according to theRMA, which, unlike many other vendors, does not include reuse of securities in theirdata (Table 2).

Initially, risk aversion due to counterparty risk immediately following Lehman ledmany pension and insurance funds’ official accounts to not let go of their collateralfor incremental returns (ie, supply was constrained). More recently, demand-sidepressures, such as the regulatory squeeze on the use of balance sheet and low returnson cash holdings, have put a lid on this market. These figures are not reboundingas per the end-of-2015 financial statements of banks. The RMA’s data includes thelargest custodians, such as BNY, State Street and JP Morgan.3

Some suggestions for uplifting the securities lending market in the new regulatoryenvironments include the following.

2 It is standard practice to use title transfer in repo and securities-lending activities. (Securitieslending transactions in the United States are done via pledge; securities lending in Europe involvestitle or “pure” transfer.) Further, with respect to legal rights, securities lending is effectively identicalto repo; however, some securities lenders take the view that their clients’ rights are more securethan they would be via a repo. This is due to the indemnification of the borrower’s potential failureto return securities or default. In Europe, the securities lending is done via the GMRA or theGMSLA. (In the United States, the respective documents are the MRA and the MSLA.) Also, OTCderivatives contracts under the ISDA use English law, in which title transfer is part of the creditsupport agreements (CSAs).3 The decline in the first row of Table 2 requires some explanation. The US regulatory rules that guideborrowers permit only cash and certain government securities (and investment grade corporates).Hence, the United States developed as a cash collateral business, where the lending agent lendsclient assets versus cash and then reinvests the cash according to the client’s direction in very short-term reinvestments. Outside of the United States (in the United Kingdom, for instance), regulatoryrules permit certain types of noncash collateral that are readily available (such as Financial TimesStock Exchange (FTSE) equities). In the aftermath of Lehman and the liquidity crisis, borrowersin the United States borrowed more hard-to-borrow stocks (specials) and less general collateral;this explains the decline. Noncash collateral deals (ie, collateral for collateral) effectively providelenders with a hard fee for the deal; however, these deals do not give temporary cash to generateexcess returns by creating a short-term, money-market book.





74 M. Singh

TABLE 2 Sources of pledged collateral, velocity and collateral (2007; 2010–13).

Collateral received from pension funds, insurers,official accounts, etc (US dollars, billions)‚ …„ ƒ

2007 2008 2009 2010 2011 2012 2013 2014 2015

Securities 1209 935 875 818 687 620 669 701 644lending vscashcollateral

Securities 486 251 270 301 370 378 338 425 454lending vsnoncashcollateral

Total 1695 1187 1146 1119 1058 998 1008 1137 1098securitieslending

All data in US trillions; velocity in units.

� The noncash collateral market in the United States should work toward those inEurope; at present, the United States has more attractive collateral rates (thanelsewhere), in part due to the repo rates being floored at 25 basis points (bps)at present, which is due to the Fed’s monetary policy.

� Equities can be increasingly mobilized and swapped with US treasuries,but regulations may need to change here (eg, the Securities and ExchangeCommission’s (SEC’s) Rule 15c3).

� Moreover, large holders of good collateral (eg, US treasuries) in the Gulfregion or some Asian countries cannot lend, as their rules prohibit the net-ting of sovereign client’s transactions (ie, their immunity angle).4 Given thehigher leverage ratio requirements for G-SIBs (especially in the United States),

4 For transactions collateralized by cash, the collateral receiver gives out cash and has a receivable(asset), and the collateral provider receives cash and books as payable (liability). Essentially, thetransaction is booked as a cash loan, or cash borrowed, collateralized by the security lent (or repo-ed) from an accounting standpoint. Under US generally accepted accounting principles (GAAP),if certain conditions are met (ie, same counterparty, same explicit maturity date (not open), intentto net settle, master netting agreement in place and legal right to offset in default), only then mayaccounts receivables and payables be netted down. Under Basel rules, if similar, but slightly moreexpansive, requirements are met, then the transactions may be netted. The legal right to offset indefault has led to many prime brokers determining that certain counterparties, most specificallySWFs and central banks, cannot be netted.






certain transactions do not make economic sense for some prime brokers toenter.5 Note that noncash trades are off balance sheet unless the collateral isre-hypothecated, so the re-hypothecation is what leads to a leverage issue.

� While the supply side (ie, central banks and SWFs) may be eager to increaselending, and the demand side (ie, hedge funds) may be eager to increase bor-rowing, the intermediaries (ie, large banks and agency lenders) will remain con-strained by the regulations for banks’ leverage and liquidity ratios; for agents,single counterparty credit limits and conservative risk-based capital rules.

If the market were to grow back to pre-crisis size, it would probably involve a muchlarger participation by nonregulated institutions, and/or connect supply to demandwithout an intermediary. While this is possible (the FSB already has a working groupto look at nonbank-to-nonbank collateral moves), it would be a very different marketfrom that which operates today, and one in which credit and duration managementand intermediation would have to be assumed by a different group of players, andpotentially under a different set of rules.

3.2 Derivative markets use of collateral

Unlike the “gross” flow of collateral in repo and securities lending, in the OTC deriva-tives market, the collateral flows in line with the risk (and, thus, on a “net” basis).Still, the “undercollaterization” in this market is large, about US$3 trillion by Bankfor International Settlements estimates (Table 3), which, if calibrated further, suggeststhat a sizable flow of collateral (or cash) will be required through the balance sheets.This may be arduous, as many of the initial margins are not allowed to move, so theywill be “parked” somewhere on a balance sheet. Collateral velocity is much lowernow than in pre-Lehman times, and if we adjust for this metric, then it is unclearwhether balance sheets have the space to accommodate the required flows, unless theregulations create balance sheet space (eg, by tweaking the leverage ratio, as acknow-ledged by the Bank of England’s Financial Policy Committee Statement minutes ofJuly 2016).6

5 Also, in the United States, almost all states allow netting, so it is easier for large pension funds/insurers to securities-lend to the large domestic banks.6 It is useful to mention a particular bias and the way it affects our regulators. In repo and securitieslending, collateral moves gross: ie, if X has $100 million exposure to a bond and needs financing, Xwill send the full $100 million market value of collateral (ie, the bond) and will receive $90C millionin funding after haircut. That makes big numbers from small risk: numbers, in fact, that are “ona par with money metrics”. In derivatives, both of those big numbers (my $100 million exposureand your $90C million exposure) are imbedded in the swap, and collateral travels for the risk only(ie, net).





76 M. Singh

Cognizant of the dilemma and push toward mandatory clearing of standard con-tracts at CCPs, these large institutions are now allowed to park client margins atcentral banks (including at the Fed).

3.3 Prime brokerage

Some of the recent prime brokerage activity indicates that equity long/short positions(ie, the delta bias) and associated netting are more balance-sheet friendly than othercollateral transactions (see Figure 2). Intuitively, the more long positions there are rel-ative to short positions, the more collateral is released to the market. Hedge funds bor-row from prime brokers (mostly the ten to fifteen banks alluded to in Box 1) for equitylong/short and event-driven (eg, credit/distressed and merger arbitrage) strategies.Since Lehman, hedge funds have financed via prime brokerage and repo strategiesroughly equally (adjusting for derivatives use/leverage within each strategy).

However, in the last couple of years, regulations have changed incentives, sinceequity/long short are “netted” on balance sheet and thus require less balance sheetspace from the prime broker (unlike repos, which are gross positions on the balancesheet). Most recent flows suggest a much higher fraction of collateral released to themarket via prime brokerage (about US$1.3 trillion) relative to about 700 billion viarepo strategies. These figures were roughly US$900 billion each via prime brokerageand repo prior to the Lehman crisis.

The accounting for prime-brokerage lending and short covering offers more oppor-tunities for balance sheet netting than are offered by other contractual forms for thesame market risk. In repo, each transfer of cash between counterparties (with limitedexceptions) is separately accounted for as an asset or liability. In prime brokerage,the customer’s net cash position after all security purchases and sales is all that goesdirectly on the balance sheet. So, if the prime broker can minimize on-balance-sheettrades with non-prime-brokerage customers that are required to meet the securities andcash needs of their prime-brokerage customers (by rehypothecating one customer’slong position to deliver against another’s short, for instance), then they can minimizetheir reported balance sheet. Simply put, the accounting in prime brokerage followsthe money, not the securities. The more the prime broker is able to optimize securitiesavailable against securities needed, the smaller the balance sheet required to providethe same services.

In summary, long/short equity via prime brokerage looks to be the best optionso far, with a “net” and elasticity of 140% (see Appendix A). This is followed byderivatives, as collateral flows on a “net” basis only; then repo, as it is primarily forfunding and not to augment returns; and then securities lending. However, some ofthe biggest clients will now be “lost”, as they are sovereign (and their immunity doesnot allow for netting).






TAB

LE

3R

isk

afte

ren

forc

emen

tofn

ettin

gag

reem

ents

inO

TC

deriv

ativ

esm

arke

t:un

der-

colla

tera

lizat

ion

inth

eO

TC

deriv

ativ

esm

arke

t.

Gro

ssm

arke

tva

lue

‚…„

ƒH

220

08H

120

09H

220

09H

120

10H

220

10H

120

11H

220

11H

120

12H

220

12H

120

13H

220

13H

120

14H

220

14H

120

15H

220

15

Gra

ndto

tal

3528

125

314

2154

224

673

2129

619

518

2728

525

392

2474

020

245

1882

517

438

2088

015

313

1449

8(a

)F

EX

cont

ract

s4

084

247

02

070

252

42

482

233

62

555

221

72

304

242

72

284

172

42

944

235

92

579

(b)

Inte

rest

rate

cont

ract

s20

087

1547

814

020

1753

314

746

1324

420

001

1911

318

833

1523

814

200

1346

115

608

1106

210

148

(c)

Equ

ity-li

nked

cont

ract

s1

112

879

708

706

648

708

679

645

605

692

700

678

615

606

495

(d)

Com

mod

ityco

ntra

cts

955

682

545

457

526

471

487

390

358

384

264

269

317

237

297

(e)

Cre

ditd

efau

ltsw

aps

511

62

987

180

11

666

135

11

345

158

61

187

848

725

653

635

593

453

421

(f)

Una

lloca

ted

392

72

817

239

81

788

154

31

414

197

71

840

179

277

972

467

180

359

655

8

Gro

sscr

edit

expo

sure

*5

005

374

43

521

357

83

480

297

13

912

366

83

626

378

43

033

282

63

358

287

02

853

*Gro

ssm

arke

tva

lues

have

been

calc

ulat

edas

the

sum

ofth

eto

tal

gros

spo

sitiv

em

arke

tva

lue

ofco

ntra

cts

and

the

abso

lute

valu

eof

the

gros

sne

gativ

em

arke

tva

lue

ofco

ntra

cts

with

non-

repo

rtin

gco

unte

rpar

ties.

Gro

sscr

edit

expo

sure

isaf

ter

taki

ngin

toac

coun

tleg

ally

enfo

rcea

ble

bila

tera

lnet

ting

agre

emen

ts.





78 M. Singh

FIGURE 2 Equity long/short hedge fund position (ie, delta bias).

0

500

1000

1500

2000

2500

0

10

20

30

40

50

60

70

Long

/sho

rt b

ias

(%)

S&

P 5

00 in

dex

Long/short bias S&P 500 index

Jan 22009

Jan 22010

Jan 22011

Jan 22012

Jan 22013

Jan 22014

Jan 22016

Jan 22015

4 COLLATERAL REUSE AND BALANCE SHEET CONSTRAINTS

As central banks unwind their balance sheets in the future, they will be careful tolet the market have possession of securities as collateral, bought via QE, since thereuse rate of these securities is outside their control. With a large balance sheet, theunwind will be over a significant period of time, and thus not a short-term conflict,as is assumed in the monetary policy literature. Further, if central banks remain partof the plumbing and take money directly from nonbanks, the financial plumbing thatrelies on such money gets rusted, as dealer banks do not receive the money flow. Thus,the dealer banks that connect the money pools and collateral pools will unwind suchconnections, thereby leading the plumbing to rust.

As alternatives to the likely dilemma of central banks’ providing balance sheetspace, can nonbanks be providers of liquidity? Long-term asset managers (ie, lifeinsurance and pension funds) and SWFs desire collateral that is of low volatility butnot necessarily highly liquid. These entities should be net providers of liquidity, eitherin the form of cash or liquid collateral. But, critically, their “need” for collateral isrelatively static (or, as providers of liquidity, they can dictate that counterparties take afixed amount). However, hedge funds, money market funds and, with new regulations,dealer banks have a dramatically shifting need for collateral and a large number ofcounterparties. Their need is for liquid collateral. So, a market for collateral could, intheory, work. Thus, the “principal” model (that embodies the banking industry) shiftsto an “agency” model. Presently, it is not possible for nonbanks such as pensions andinsurers to directly deal with other nonbanks such as hedge funds, since the latterare not rated; such constraints will keep the global banks at the center of financialplumbing (unless replaced by central banks).






FIGURE 3 Plumbing with both private and public balance sheets.

Source: author’s illustration.

5 CONCLUSION

QE created excess reserves, but removing them from the financial system impactselements of plumbing that will need to be incorporated into monetary policy decisionmaking. The new regulations that constrain bank balance sheets further impede marketplumbing. However, given the role of the banking system as conduits for collateralflow, the plumbing will always be available for privileged clients of the banks (orcustodian banks) but not for everyone else, since the private balance sheet space isbeing rationed. Going forward, the choice of balance sheet, private or public, shouldbe driven by market forces, and not by the ad hoc allocation of central banks (seeFigure 3, where the red area is the reduction in plumbing since money flows directlyto the central bank and not to the market plumbing (or blue area)). More importantly,monetary policy transmission is weakened if parts of the plumbing move to a centralbank balance sheet.

APPENDIX A. REHYPOTHECATION AND THE LEHMAN EPISODE

Since the Lehman Brothers bankruptcy, there has been criticism from the UnitedStates that the United Kingdom does not have rigid quantitative regulatory caps onrehypothecation equivalent to those applicable to broker-dealers regulated by theSEC in the United States (even though many UK brokers agree to caps in contracts).





80 M. Singh

BOX 2 The global bilateral collateral market (relative to the US triparty repo market).

Collateral use and reuse in financial markets is popular.Before the Lehman crash, the volumeof funding via pledged collateral (including title transfer) was about US$10 trillion, higher thanthe US broad measure of money, M2.This box tries to summarize the difference between themuch-researched triparty “repo” market and the less-researched bilateral collateral market;the latter includes collateral flows from not only bilateral repo but also securities lending,derivatives and prime brokerage.

The US bilateral repo market is a subset of the “market for collateral” (securities for posses-sion and use, incidentally against cash). The triparty repo (TPR) market in the United Statesis a “market for funding” (money for broker-dealers/banks, incidentally collateralized by secu-rities). The TPR market is currently estimated at US$1.6 trillion from a peak of almost US$3trillion before the Lehman crisis. The TPR market provides banks with cash on a securedbasis, with the collateral being posted to cash lenders (eg, money market funds) through oneof the two clearing banks: BNY Mellon and JP Morgan. The bilateral repo market is sizable,and although no official statistics exist, some recent work at central banks suggests thismarket to be on par with or bigger than the TPR market (eg, the New York Fed estimates thismarket to be between US$1 and US$2 trillion in the United States alone) (Baklanova et al2016).

Think of the bilateral repo market using this analogy for the old-clothing trade.Typically, mer-chants in developed countries shrink wrap old clothes in shipping-container-sized bundles(under pressure) and send these plastic-wrapped blocks to poor countries. There, clothingbrokers buy the blocks and resell them by weight to jobbers. So, if a block weighs 500 pounds,and the broker decides to sell it in 10-pound lots, fifty people will gather around to make apurchase. However, some people will pay slightly more to be at the front of the crowd, andsome will pay slightly less to be at the back. When the jobber pops the bundle open with abig knife and the shrink wrap explodes, everyone gathered around scrambles for the bestpieces. Collateral desks are a bit like those jobbers. Big lots come in from hedge funds andsecurity lenders, and the large bank’s collateral desk paws through it, searching for gems.Those gems go out bilaterally to customers who will pay a premium. The remainder goesto the guys at the back of the crowd (for example, TPR repo). To the extent that securitieseligible for the TPR market are in demand in the bilateral market, banks will generally usethem first in the bilateral market, as it offers a better price.

The figures shown above that depict the bilateral-pledged collateral do not count TPR-relatedcollateral, as this is trapped within the TPR structure. The operational structure of the RRPfacility puts practical restrictions on the reuse of collateral outside the triparty system. Col-lateral can only be used in a triparty repo liability. (So, a firm that is a “dealer” in the tripartysystem, such as JPMorgan Chase or BNY Mellon, could have as an asset a Fed RRP and asa liability a triparty repo with a customer.) Members of the Government Securities Division(GSD) of the Depository Trust and Clearing Corporation (DTCC) can reuse the collateralwithin the General Collateral Finance (GCF) triparty system. Here, we use the term “banks”very loosely; for example, Citibank could take collateral from the Fed and give this to a Fidelitymutual fund as a triparty investment, or it could take collateral from the Fed and give this inthe GCF to Credit Suisse to give to that Fidelity fund. To be clear, members of the GSD maybe classified differently: Goldman Sachs is actually Goldman Sachs & Co., Deutsche Bankis Deutsche Bank Securities Inc. and Barclays is Barclays Capital Inc. Members also includePierpont Securities LLC, Jefferies LLC and Cantor Fitzgerald & Co. The important point isthat reuse of collateral can only end in a triparty repo; it can have no other use outside thissystem.






Specifically, some feel that this asymmetry is akin to regulatory arbitrage and that theUnited Kingdom offers a unique forum for “unlimited rehypothecation”.7

Proposed regulations seem to be at odds with “title transfer”. If I transfer title,then the recipient of collateral is able to use that asset in any way they deem fit.This is not compatible with regulations that treat the asset as “client property” andlimit rehypothecation or segregate for the client. In fact, insisting on segregationundermines the legal construction under which title was transferred. An importantdistinction is the interpretation of the prefix “re” in “rehypothecation”. In the UnitedStates, this is normally done with a pledge with consent to reuse. So, there is a cleardistinction between pledged securities and sold securities. However, in Europe a repois a contract of sale with a promise to repurchase at an agreed future date and price.Legally, if I sell securities, the resulting securities are no longer my securities; if thesesecurities are then onward pledged, that is not a rehypothecation from my angle.However, is this economically different if I sell securities on the basis that you agreeto sell me equivalent securities at some future time? The Basel approach is along thelines that the existence of the promise to sell back means that the original sale is nolonger a “pure” sale, and therefore it is caught by the rehypothecation restrictions.

But these criticisms risk overlooking three significant counterarguments. First, assubsequent litigation has revealed, the UK broker Lehman Brothers InternationalEurope (LBIE) appeared to have broken the UK rules on client asset segregation. Incertain cases, it appears that LBIE had not been properly segregating client property.Quantitative limits on reuse do not protect clients whose brokers do not follow therules. Second, it could be argued that Lehman clients who had voluntarily agreed togive broad rights of reuse in their prime-brokerage contracts essentially got what theybargained for when LBIE failed. Those clients (for the most part, professional andsophisticated counterparties) had misjudged the counterparty credit risk on Lehman,but they had not been cheated any more than an uninsured depositor is “cheated”by a failing bank. Third, the supposed uniqueness of the UK legal regime is perhaps

7 A key reason why hedge funds may have previously opted for funding in Europe is that leverageis not capped as in the United States via the 140% rule under Rule 15c3–3. In the United States,the SEC’s Rule 15c3–3 prevents a broker-dealer from using its customer’s securities to financeits proprietary activities. Under this rule, the broker-dealer may use/rehypothecate an amount upto 140% of the customer’s debit balance (ie, borrowing from the broker-dealer). As an example,assume a customer has US$500 in pledged securities and a debit balance of US$200, resulting innet equity of US$300. The broker-dealer can rehypothecate up to US$280 of the client’s assets(140% � US$200). Created by the Securities Investor Protection Act (SIPA), the Securities InvestorProtection Corporation (SIPC) is an important part of the overall system of investor protection inthe United States. SIPC’s focus is very specific: restoring securities (rather than cash) to investorswith assets in the hands of bankrupt brokerage firms (eg, Lehman). MF Global is a useful recentprecedent.





82 M. Singh

overplayed: the types of counterparties that go to London, rather than Frankfurt orParis, do not do so for any unique features of UK law. In fact, the strong legal basisfor title-transfer financial collateral actually has its roots in English law, which alsounderpins the Financial Collateral Directive of the EU. The market is in London notbecause it offers unique arbitrage but because UK courts are viewed as having a longhistory of contractual adjudication and legal principles.


The authors report no conflicts of interest. The authors alone are responsible for thecontent and writing of the paper. The views expressed are done in a personal capacityand should not be reported as representing the views of the IMF or IMF policy.

REFERENCES

Baklanova, V., Caglio, C., Cipriani, M., and Copeland, A. (2016). The US bilateral repomarket: lessons from a new survey. Report, Office of Financial Research.

European Systemic Risk Board (2014). An analysis of the ESRB’s first data collection onsecurities financing transactions and collateral (re)use. Occasional Paper 6, September,ESRB.

Gourinchas, P.-O., and Jeanne, O. (2012). Global safe assets. Working Paper 399,December, Bank for International Settlements.

Singh, M. (2011).Velocity of pledged collateral: policy and analysis.Working Paper 11/256,International Monetary Fund.

Singh, M. (2014). Collateral and Financial Plumbing. Risk Books, London. URL: http://bit.ly/2aGBgcF.

Singh, M. (2015). Managing the Fed’s liftoff and transmission of monetary policy. WorkingPaper 15/202, International Monetary Fund.






Research Paper

Impact of monetary policy on collateral reuse

Ameya Muley

Massachusetts Institute of Technology, 77 Massachusetts Ave, Cambridge, MA 02139, USA;email: [email protected]


ABSTRACT

Rehypothecation is the direct reuse of collateral received by a lender to borrow ontheir own account from a third party. It enables the intermediate lender to maximizeinvestment in their profitable project by making efficient use of the collateral. I showthat when a central bank removes collateral from the market through an open marketoperation, less collateral is available to be borrowed by the productive intermediatelenders, leading to retarded investment by them and lower aggregate output. This isdue to a pecuniary externality: the presence of other competing lenders increases thecost of rehypothecation for the productive lenders to inefficient levels. I find empiricalevidence for this channel in the cost of borrowing in the bilateral repurchase agreement(repo) market. The policy implications of this result include conducting open marketoperations when collateral is abundant and the repo rate is high. It also suggests thatusing interest on reserves may be more effective as a policy tool compared with openmarket operations when collateral is scarce.

Keywords: collateral reuse; rehypothecation; monetary policy; open market operations; quantita-tive easing.

Print ISSN 2049-5404 j Online ISSN 2049-5412Copyright © 2016 Incisive Risk Information (IP) Limited

83




84 A. Muley

1 INTRODUCTION

The spectacular growth in shadow banking over the past couple of decades (Cetorelliet al 2012; Pozsar et al 2012) has led to a significant increase in the demand for and uti-lization of financial securities as collateral in short-term repurchase agreements. Thelimited supply of “safe” collateral has also led to an increase in the practice of reusingavailable collateral. Rehypothecation is the direct reuse by lenders of the collateralposted with them to borrow on their own account. This is commonly observed inprime-brokerage arrangements, where prime brokers rehypothecate their hedge fundclients’ collateral to borrow from money market funds (MMFs), and in repo markets,where cash lenders reuse the securities they receive in a repurchase agreement. Datafrom 10-Q filings with the Securities and Exchange Commission (SEC) shows thatin 2013, approximately US$2 trillion dollars of collateral was rehypothecated by theUS broker-dealers, or about 30% of their total assets. The rapid expansion in shadowbanking has also made it necessary to understand the transmission of monetary policythrough these funding channels characterized by extensive collateral reuse.

In this paper, I provide theoretical microfoundations to understand the impact ofmonetary policy on markets characterized by collateral reuse. I study a model of inter-mediation through rehypothecation with the repurchase agreement (repo) and primebrokerage markets in mind. The model is motivated by the commonly observed fund-ing chain in the prime-brokerage market, where hedge funds (the original borrowers)borrow from their prime brokers (the intermediate lenders), who rehypothecate thecollateral to MMFs (the ultimate lenders). In my setup, both the borrower and theintermediate lender have variable scale investment projects. I show that rehypotheca-tion allows the intermediate lender to maximize investment in their profitable projectby lending cash at a high haircut, and then rehypothecating the collateral at a lowerhaircut. The intermediate lender, thus, effectively borrows the haircut from the orig-inal borrower. The extent of this borrowing is restricted by the pledgeability of theintermediate lender. Pledgeability is the upper bound on borrowing, as a fraction of aninvestor’s expected cashflow, that does not trigger incentive problems. The higher thepledgeability, the stronger the ability of the intermediate lender to effectively borrowfrom the borrower at high haircuts.

Next, I consider the central bank’s activities as a way of providing an alternativesource of funding to the hedge funds in this segmented market. Following the quan-titative easing actions of the Federal Reserve (Fed), concerns have been raised bymarket participants about the “illiquidity” in the treasury market. Singh (2013) hasalso documented a fall in the “velocity” (the number of times a piece of collateral ispledged and repledged) of treasury collateral after QE2 and QE3. Motivated by this,I use the optimal rehypothecation contract to study the intervention of a central bankin the repo market and show that the central bank’s actions may be ineffective. The





Impact of monetary policy on collateral reuse 85

collateral removed from the system sits with the central bank and does no work. Asa result, the effectiveness of the expansionary operation is reduced, and if the inter-mediary’s project is more valuable socially than the borrower’s project, it leads to afall in aggregate output. The reason for this is a pecuniary externality. The borrowerin the rehypothecation chain gives up the collateral to the central bank, raising thecost of borrowing collateral for the intermediary to inefficiently high levels. Due tolimited pledgeability, the intermediary cannot outbid the central bank to borrow theefficient amount of collateral. This moves investment away from the intermediary’sproject toward relatively less valuable projects, leading to a fall in the aggregate out-put. Conversely, a contractionary operation has the side effect of stimulating privatelending. This is because the collateral that was hitherto locked up with the centralbank is released and loosens borrowing constraints.

This effect holds when collateral is scarce and the cost of borrowing it in the privatemarket is high. When collateral is abundant, however, some of it is idle and is not beingused for borrowing and rehypothecation. In this case, the central bank’s interventionsimply takes away the idle collateral without affecting rehypothecation and investmentdown the funding chain. I suggest the bilateral repo spread as a measure of collateralscarcity. When collateral is idle, the cost of borrowing it in the private market is zero,and the repo spread is high. When collateral is scarce, the cost of borrowing it is highand the repo spread is low. I find evidence for this channel in the general collateralbilateral repo spreads for US treasury securities. When the repo spread is high (as itwas before 2009), an increase in the supply of treasuries has little effect on the bilateralrepo spread, indicating that there is little effect on the cost of rehypothecation downthe chain. When the repo spread is low (as it has been since 2009), an increase inthe supply of treasuries significantly increases the repo spread, indicating a fall in thecost of rehypothecation.

This has a number of policy implications. The result shows that open market opera-tions must be conducted when collateral is abundant and idle. In view of the imminentreversal of many of the world’s central banks’ quantitative easing policies, too, it hasimportant implications. First, it suggests that tools such as interest on reserves, whichdo not affect the supply of collateral, would be more effective than open market oper-ations when collateral is scarce. Further, high quality collateral such as treasuriestends to be relatively scarce in times of crisis. Conducting expansionary open marketoperations using this safe collateral at precisely these times is more likely to back-fire. Second, the result indicates that when the central bank conducts expansionaryopen market operations, it should restrict the purchases of assets from dealers andmoney funds to directly provide liquidity to where it is most valuable. Also, whenthe central bank, with a contractionary intent, borrows cash against collateral throughfacilities such as the overnight reverse repurchase agreement (ON-RRP), it should





86 A. Muley

deny rehypothecation permissions to the counterparties receiving the collateral, asrehypothecating the collateral may generate additional financing activity.

1.1 Related literature

There has been a host of recent papers focussing specifically on rehypothecation.Bottazzi et al (2012) study the existence of equilibria with limited and unlimitedrehypothecation, but they abstract from considerations of the lender’s default risk. Intheir recent papers, Eren (2014) and Infante (2014) specifically model rehypotheca-tion and view it as a way for the intermediate broker to obtain liquidity by offeringdifferential haircuts: higher haircuts to the cash borrower and lower haircuts to thelender on rehypothecation. Maurin (2014) considers a general equilibrium model withcollateral constraints and rehypothecation in a frictionless setting and finds that rehy-pothecation can at best be a substitute for complete markets. Andolfatto et al (2015)focus on the liquidity creating role of rehypothecation and argue that limiting it maybe desirable in increasing the demand for cash balances in economies away from theFriedman rule. Lee (2015) studies the effect of collateral reuse on repo spreads.

The results on the effect of central bank intervention on repo markets relate to sev-eral papers that compare the different tools of monetary policy. Goodfriend (2002)suggests that the market interest rate and the level of reserves can be set indepen-dently, while Keister et al (2008) describe a floor system to divorce the two. Martinet al (2013) analyze the effectiveness of the various tools of the Fed in maintainingthe floor on rates. Stein (2012) prescribes the use of interest on reserves to avoidexcessive short-term debt creation, while Kashyap and Stein (2012) advocate theuse of a combination of open market operations, reserve requirements depending onthe nature of the financial system. Cochrane (2014) suggests that a regime with alarge central bank balance sheet and interest paying reserves creates financial sta-bility, as the interest paying reserves reduce the incentives to create run-susceptible“inside money”. Ewerhart and Tapking (2008) study the optimal choice of collateralin repurchase agreements and examine the welfare implications of the central bank’spurchases of different types of collateral. Araújo et al (2013) study the effect of openmarket operations on borrowing constraints. While, in their paper, asset purchasesmay tighten or relax collateral constraints depending on the risk of the asset, in mymodel, asset purchases tighten collateral constraints due to an externality.

The effect of treasury supply on their prices is investigated by Krishnamurthy andVissing-Jorgensen (2012), who find that treasury securities enjoy a scarcity premium.Krishnamurthy and Vissing-Jorgensen (2013) argue that depriving the economy ofliquid treasury bonds may have adverse welfare consequences. D’Amico et al (2013)find that purchases of specific Committee on Uniform Security Identification Pro-cedures (CUSIPs) by the Fed leads to an increase in the corresponding special repo






spread as the supply decreases. Fleming et al (2010) similarly find that the repospread between treasury and non-treasury collateral narrowed due to the Term Secu-rities Lending Facility (TSLF), which lent treasuries against other assets. Caballero(2006) and Caballero et al (2006) discuss the macroeconomic and international impli-cations of collateral shortages, and Caballero and Farhi (2013) highlight the benefitsof supplying safe assets to the system.

1.2 Outline

Section 2 describes the baseline model, the contracts and the timing as well as theoptimal rehypothecation contracts. Section 3 looks at a central bank intervention ina market characterized by rehypothecation and the empirical evidence and policyimplications. Section 4 concludes.

2 BASELINE MODEL

I describe here a model of rehypothecation of collateral. There are two dates, t D 1; 2.The first period is further divided into two: t D 1:1 (beginning of the period) andt D 1:2 (end of the period). There are three types of agents: A, B and C. Thereis a continuum of unit mass of each type. I interpret an agent of type A to be ahedge fund that needs to borrow, and an agent of type B to be an investment bank orprime-broker that can lend to the hedge fund A against collateral, and that can laterrehypothecate that collateral. Agent C is interpreted as an MMF that lends to bank Bagainst collateral.

There is one good called cash and an asset called collateral. Cash is the numeraire.Agents of type A, B and C are risk neutral and consume cash at t D 2. There is nodiscounting. One unit of the collateral pays off 1 unit of cash at t D 2, with probability1. At t D 1:1, A has ˝ units of the collateral asset, and B has 1 unit of cash. C has alarge quantity of cash at each period that can be stored at an interest rate of 0 betweenany two periods. At t D 1:1, A has access to a variable scale investment opportunitythat yields RA > 1 units of cash, with certainty at t D 2 per unit investment of cash.At t D 1:2, B has access to an investment opportunity that yields RB > RA units ofcash at t D 2 with certainty. B cannot store the cash between t D 1:1 and t D 1:2. Bhas access to a storage technology with interest rate 0 between t D 1:1 and t D 2.

2.1 Borrowing contracts

To finance their investments, A must borrow from B, and B must borrow from C.The contract between A and B can be thought of as a prime-brokerage contract or a





88 A. Muley

FIGURE 1 Borrowing contracts.

A B C(r,h,θ) ˜ ˜(r,h)

The arrows indicate the flow of cash at the time of lending.

bilateral repo contract, whereas the contracts with the MMF C can be thought of as atri-party repo contract.1

Figure 1 describes the contracts. I assume without loss of generality that thesecontracts are offered by B to A and C. I first consider the contracts CA;B D .r; h; �/

between A and B, traded at t D 1:1. Here, r is the net interest rate, h is the haircut,2

and � 2 f0; 1g is the rehypothecation permission: B can rehypothecate the collateralif � D 1 and not otherwise.3 At t D 1:2, B may borrow from C under the contractCB;C D . Qr; Qh/, depending on what contracts were traded between A and B at t D 1:1.Here, Qr is the interest rate, and Qh is the haircut. If � D 1, B can directly rehypothecatethe received collateral asset. If � D 0, B can only hold on to the debt. I assume thatthe borrowers prefer borrowing against the collateral to selling the collateral, and thatthe collateral provided in the contract is of value at least equal to the face value of thedebt, ie,4

1 C r 6 1

1 � hand 1 C Qr 6 1

1 � Qh; (2.1)

where, in each inequality, the left-hand side is the face value of the debt per unitof borrowing, and the right-hand side is the value of collateral pledged per unit ofborrowing.

At t D 2, the returns to A’s and B’s investments are realized, CB;C is settled firstand, finally, CA;B is settled. When CA;B is being settled, B must return to A collateralof value equal to or greater than the face value of the debt. What B owes to A is morethan what A owes to B, and, consequently, A is an unsecured creditor of B. Following

1 Bilateral repos are traded over the counter between hedge funds and broker-dealers as well asbetween two broker-dealers. Money market funds enter into repo contracts with broker-dealers inan arrangement intermediated by JP Morgan or Bank of New York Mellon, who hold the collateralon behalf of the MMFs. This is called a tri-party repo.2 Haircut D 1 � units of cash borrowed=units of collateral pledged.3 The binary permission is without loss of generality. The framework does not preclude B offeringmultiple contracts to A. Some of these contracts may have � D 1, and others may have � D 0,so that, in aggregate, any fraction between 0 and 1 of the total collateral may be permitted to berehypothecated.4 Muley (2016) shows how the borrowers relative optimism about the collateral’s payoff leads tothis realistic result.






Holmstrom and Tirole (1997), I assume that B’s effective borrowing from A cannotbe greater than a certain fraction � of the income from B’s investment; otherwise, thehigh levels of B’s effective debt to A will trigger incentive problems. Thus, � is B’spledgeability to A. This idea is formalized in the appendix (available online).

2.2 Optimal allocations

Figure 2 summarizes the optimal contract CA;B in equilibrium. The vertical axisrepresents the net return NR to A per unit of collateral lent to B, where

NR D .RA � .1 C r//.1 � h/: (2.2)

This is the net return that A gets by borrowing from B against one unit of collateraland investing the proceeds in their project. The blue curve represents the supply zA

A;B

of collateral by A as a function of NR. A would like to allocate all of its collateral toCA;B if it gets a positive net return, and they are indifferent if the net return is zero.Hence, zA

A;B takes any value between zero and ˝ for NR D 0, and it is ˝ for NR > 0.The red curve represents B’s demand zB

A;B for rehypothecating collateral, as a functionof NR. This demand is downward sloping because when NR is high, A demands a highnet return per unit of collateral lent. This translates into a low interest rate and haircut,and therefore a diminished ability of B to demand and rehypothecate collateral. ForNR > NR�, the net return demanded by A is too high, and the interest rate that A is

willing to pay to B is too low for B to find it profitable to lend cash to A; thus, B’sdemand for collateral falls to zero.5

Let the total cash investments in A’s and B’s projects be IA and IB, respectively.Since zA

A;B is the collateral supplied byA in equilibrium, IA D zAA;B.1�h/ D zB

A;B.1�h/. Now, C will lend to B at a haircut and interest rate of zero, since C has a largeamount of cash that it can store at an interest rate of zero. Hence, by rehypothecatingthe collateral received in the contract CA;B, B is able to invest IB D zB

A;B. The followingproposition summarizes the equilibrium investments.

Proposition 2.1 (Rehypothecation allocations) In equilibrium,

(1) IA D 1, IB D RA=.1 � �RB/ and NR D 0, if ˝ 2 ŒRA=.1 � �RB/; 1/,

(2) IA D 1, IB D ˝ and NR 2 Œ0; NR��, if ˝ 2 Œ1=.RB/; RA=.1 � �RB/�,

(3) IA D ˝RB, IB D ˝ and NR D NR�, if ˝ 2 Œ0; .1=RB/�.

5 For NR < RB � 1, B’s pledgeability constraint binds and there is overcollateralization, ie, 1 C r <

1=.1 � h/. For NR > RB � 1, A’s collateral constraint binds and there is no overcollateralization, ie,1 C r D 1=.1 � h/. The proofs are included in the appendix (available online).





90 A. Muley

FIGURE 2 Contract CA;B in equilibrium.

1

E

R–

R*–

RB – 1

RB 1 – κRB

RAΩ

–zA,B(R)B

–zA,B(R)A

From Figure 2, it can be seen that when ˝ > RA=.1 � �RB/, B’s borrowing ofcollateral from A is limited by B’s pledgeability. B lends all of their cash endowmentto A, but they are able to rehypothecate only a part of A’s collateral. The remainingcollateral lies idle with A. Here, A has abundant collateral and gets a zero net returnNR D 0. When ˝ < 1=RB, A’s borrowing of cash from B is limited by the endowment

of collateral. B lends only a part of their cash endowment to A and rehypothecatesall of A’s collateral. Here, A has very scarce collateral and gets a high net returnof NR D NR�. For 1=RB 6 ˝ 6 RA=.1 � �RB/, B lends all of their cash to A andrehypothecates all of A’s collateral.

The aggregate expected output at t D 2 is given by (modulo constants)

Y D .RA � 1/IA C .RB � 1/IB: (2.3)

It can be shown that the optimal rehypothecation contract maximizes the aggregateexpected output, subject to collateral and pledgeability constraints.

3 CENTRAL BANK INTERVENTION

I now look at what happens when a central bank intervenes in an existing chain toexchange cash for collateral. In addition to the model in Section 2 with the observedmarket segmentation, I introduce a central bank that can lend directly to agent A,






FIGURE 3 Expansionary operation: central bank lending against collateral.

A B C

CB

(r,h,θ) ˜ ˜(r,h)

(rCB,hCB)

thus providing an alternative to the contract CA;C. I show that a monetary policyaction of the central bank may lose bite, as the taking out or putting in of col-lateral in exchange for cash will affect collateral constraints down the chain in acounterproductive way.

The channel I highlight here works through changing repo rates and quantities oftreasury collateral available with borrowers to pledge to private lenders. It is inde-pendent of the monetary policy channels that work through changing bond prices andinterest rates. Although this channel will be operating in normal situations, a usefulway to think of the model is in a situation in which large-scale purchases or sales oftreasury collateral are carried out by the central bank. Such purchases were under-taken by central banks in advanced economies in response to the global financialcrisis. There is also a possibility of large-scale sales of collateral being carried out inthe near future as these economies emerge out of the recession. The large quantitiesinvolved make this channel quantitatively important.

I assume a repo-style intervention by the central bank: the central bank borrows orlends cash against collateral, rather than buying or selling it. This is in keeping withthe tenor of the model, and does not cause a loss of generality. Purchases and sales ofcollateral will have the same effect on collateral constraints down a rehypothecationchain as the collateral enters or leaves a central bank. I also assume that the centralbank lends cash to or borrows cash from A against collateral. To recap, A can bethought of as a hedge fund or a dealer, B can be thought of as another dealer, and Ccan be thought of as a money fund. Figure 3 shows the flow of cash and collateral asthe central bank lends cash to A against collateral. The choice of A as the point ofentry has two bases. First, the unintended effects on collateral constraints arise whenthe entry is at A, and not when it is at B or C. Second, this is quantitatively relevant:Carpenter et al (2015) show that hedge funds tend to be the largest buyers and sellers oftreasury securities, far larger than broker-dealers. Their analysis estimates that of theUS$600 billion of treasuries purchased by the Fed during the second large-scale asset





92 A. Muley

purchase (LSAP-2) program, about 60% were sold by hedge funds.6 They propose apreferred habitat explanation for this evidence.

I assume that the central bank targets a supply of cash m to be infused into thesystem. (I abstract from the reasons why the bank would target a particular m.) Thisis without loss of generality. In the model, targeting a money supply is equivalent totargeting the interest rate between A and B. Before the market between A and B opensat t D 1:1, the central bank at t D 0 lends m 2 R units of cash to A until t D 2, at theterms .rCB; hCB/ set by the market in equilibrium. m > 0 indicates an expansionaryoperation and m < 0 indicates a contractionary operation. I assume that the centralbank makes the loan fully secured with 1CrCB D 1=.1�hCB/, so that only rCB is leftto be determined by the market equilibrium. The central bank rebates lump-sum allprofits to A, B or C at t D 2. As A borrows from the central bank, it moves collateralaway from B by an amount

�˝ D m1

1 � hCB: (3.1)

The changed supply of collateral affects the equilibrium between A and B. If m >

0, the central bank takes away collateral, increases NR and reduces the interest ratebetween A and B. If m < 0, the central bank supplies collateral, decreases NR andincreases the interest rate. In the new equilibrium, A is indifferent between borrowingfrom B and borrowing from the central bank, so that

.RA � .1 C rCB//.1 � hCB/ D NR: (3.2)

In the equilibrium with central bank intervention, the aggregate outcome is affectedas follows.

Proposition 3.1 In the equilibrium with central bank intervention, the marginaleffect due to the intervention of m 2 R units of cash is given by

@Y

@m

ˇ̌ˇ̌mD0

D

8̂ˆ̂̂<ˆ̂̂:̂

.RA � 1/ if NR D 0;

.RA � 1/ � .RB � 1/1

1 � hCBif NR 2 .0; NR�/;

.RA � 1/ � .RB � 1/1

1 � hCB� .RA � 1/.1 � h/

1

1 � hCBif NR D NR�:

If RB > RA, there exists an NR# > 0 such that

@Y

@m

ˇ̌ˇ̌mD0

< 0 for NR 2 .0; NR#/:

6 While hedge funds are not counterparties to the Fed’s open market operations, they can sellsecurities to the Fed through their broker-dealers, who are.






FIGURE 4 Equilibrium with an expansionary open market operation.

E

E

E

R–

R*–

RB – 1

R–

R*–

RB – 1

Ω

Ω

Ω

–zA,B(R)B

–zA,B(R)B

–zA,B(R)B

–zA,B(R)A

–zA,B(R)A

–zA,B(R)A

(a)

(c)

R–

R*–

RB – 1

(b)

(a) NR D 0. (b) NR 2 .0; NR�/. (c) NR D NR�.

The proposition says that the action by the central bank may have unintendedconsequences that work against the central bank’s objective. For example, in thecase of an expansionary operation with m > 0, the decreased supply of collateralmay tighten the collateral constraint between B and C, having an adverse effect onoutput. Figure 4 shows the equilibria in the three regions which react varyingly to anexpansionary operation with m > 0.

In the first region, when collateral is abundant and NR D 0, the action of the centralbank will have the desired effect of increasing output as A invests the extra cash.





94 A. Muley

There will be no effect on the equilibrium between A and B, as A will use the idlecollateral to borrow from the central bank.

However, when collateral is scarce in the second region and NR > 0, useful collateralis taken away from B through a lower haircut, and B cannot borrow as before byrehypothecating the collateral to C. This effect is particular to a rehypothecationchain and would be absent if the central bank had entered at B. It is clear that thetightening of B’s collateral constraint causes the monetary policy action to lose itsbite. In fact, if B’s project is more valuable thanA’s, it could possibly have the oppositeeffect and cause aggregate expected output to decrease.

The reason why output may even fall due to the expansionary intervention is apecuniary externality. A may end up giving too much collateral to the central bank,without taking into account the adverse effects of doing so on the collateral constraintbetween B and C. The central bank is only contracting withA, andA cannot internalizethe loss due to the tightening of B’s collateral constraint. The collateral sits idle withthe central bank, when otherwise it could have been used profitably by B. Eventhough the collateral is more useful to B, B cannot outbid the central bank to borrowit due to limited pledgeability � < 1. The intervention by the central bank movesinvestment away from B’s project toA’s project. To the extent that the broker-dealer B’sintermediation activities are socially more valuable than the hedge fundA’s speculativeactivities, this will adversely affect the aggregate output of the financial intermediationchain.

In the third region, NR D NR�. B is not lending its full endowment of cash to A, andthe central bank’s entry causesA to simply substitute borrowing from B for borrowingfrom the central bank. A’s borrowing and investment will increase by the differencein the haircuts h and hCB offered by B and the central bank, respectively, and B willconsume the idle cash. This effect is not particular to a rehypothecation chain; it wouldhave existed even if the central bank had entered at B.

The case of contractionary operations is nearly symmetric, with an additional fea-ture: the central bank must decide whether to allow A to rehypothecate the receivedcollateral. With a sale of the collateral, the rehypothecation permission in implicit.With lending collateral, however, the permission needs to be made explicit.

Proposition 3.2 (Contractionary operation) The marginal effect due to a contrac-tionary operation .m < 0/ is given by Proposition 3.1 if rehypothecation by A ispermitted, and by @Y=@mjmD0 D RA � 1 for all NR 2 Œ0; NR�� if rehypothecation byA is not permitted.

If collateral is not permitted to be rehypothecated by A, it just sits with A and doesnot circulate to generate more borrowing and investment.






3.1 Policy implications

The trade-off between cash and collateral created by open market operations leads toseveral policy implications. The fact that the removal of collateral during an expan-sionary open market operation can contract financing and investment suggests that theideal way to conduct such operations is for the central bank to not demand collateralfor the cash it supplies. The potential losses to the central bank due to default maythen be covered by increased taxes on the profits of the financial sector. Similarly,the effective way to conduct contractionary operations would be to not provide col-lateral in exchange for cash, and to use instruments such as reserve requirements andinterest on reserves or excess reserves (IOR, IEOR). Other tools such as the TermDeposit Facility (TDF) could be expanded, as could the coverage of reserve require-ments across institutions and liability types, in order for reserves to be effectivelymanipulated on a large scale using IOR without resorting to open market operations.Indeed, good-quality collateral such as treasuries is most likely to be scarce, and themagnitude of the pecuniary externality in likely to be strongest, in times of crisis.Conducting expansionary open market operations using treasuries at such times ismore likely to backfire.

If open market operations are to be conducted, they must be conducted in a waythat does not affect the supply of collateral that can be profitably rehypothecated. Onepotential variable to look at would be the spread between the collateralized borrowingrate and the short-term risk-free rate. A high value of this spread would mean thatcollateral is abundant and idle and that it is safe to conduct an open market operation.With contractionary open market operations, it is much easier to ensure that thecollateral provided does not enter into circulation by denying rehypothecation rights.The Fed’s ON-RRP program is conducted through a tri-party arrangement wherein thecollateral stays with the tri-party bank and is not rehypothecated. It may be tempting toconsider statutorily limiting rehypothecation to provide greater control to the centralbank over the outcomes of its monetary policy actions. However, this may not beideal, since it will prevent collateral from flowing to where it is needed the most.

Another way of minimizing the unintended consequences of open market opera-tions is to conduct them with counterparties who are at the lending end of the chain.Contractionary operations that lend collateral to money funds will be more effectivesince they are not natural rehypothecators and will hold on to it. The ON-RRP’sexpanded counterparties include such entities. In fact, primary dealers rarely avail ofthis facility, and most of the volumes are transacted with MMFs. Similarly, expan-sionary open market operations in crises can be restricted to purchase assets from themoney funds and dealers (agents B and C in the model) to ensure that liquidity isprovided directly to where it is most valuable for intermediation purposes, instead ofto hedge funds (agents A) that engage in speculative activities.





96 A. Muley

3.2 Empirical evidence

The model predicts that an increase in the supply of collateral lowers the return NR fromlending collateral and increases the repo rate r in the bilateral repo market betweenA and B.7 I consider the weekly regression

�st D ˛ C �1�st�1 C �2st�1 C X 0t� C ˇ1�Tt C ˇ2It�1�Tt C �t : (3.3)

Here, st D rt � y3mot is the overnight borrowing rate with treasury collateral, nor-

malized by the short-term risk-free rate, the three-month treasury yield. Xt are vari-ous market controls including the lagged equity market, the Chicago Board OptionsExchange Volatility Index (VIX), short- and long-term treasury yields and theirchanges. �Tt are the weekly changes in outstanding treasury securities. The productIt�1�Tt of a measure of collateral idleness and the change in treasury supply isincluded to test for the hypothesis that the interest rate reacts less to changing collat-eral supply when collateral is idle. For the index of collateral idleness, I use It D st

itself, since the model predicts that when collateral is abundant and idle, the interestrate is high. The model predicts that ˇ1 > 0 and ˇ2 < 0.

3.3 Data

The series for Tt are obtained by subtracting the treasury holdings of the System OpenMarket Account of the Federal Reserve Bank of New York from the federal debt heldby the public.8 The series for rt is the index of overnight US treasury collateral reporates, USRG1T, obtained from Bloomberg. The data is weekly from 2005 Q2 to2013 Q1. I use high-frequency weekly data to minimize lower frequency fluctuationsand other secular trends caused by regulatory changes. The weekly frequency alsoeliminates the possibility of the left-hand side variables affecting the changes in thesupply of treasuries, which are announced well in advance. I use overnight financingdata to increase the power of the test, since it is overnight financing (which needsto be rolled over every night) that will be impacted most by the changing supply ofcollateral.

7 While in the simple model with homogeneous agents, NR does not change for small changes in thesupply of collateral when NR 2 f NR�; RB � 1g, it can be shown that for a continuum of agents Bwith project returns in some interval ŒRB;L; RB;U�, @ NR=@m < 0 and @r=@m > 0 when collateral isscarce and NR > 0, and @ NR=@m D @r=@m D 0 when collateral is idle and NR D 0.8 The weekly data for the federal debt held by the public is obtained from www.treasurydirect.gov/.This data is the face value of all US treasury securities. The ideal variable to look at would bethe market value of all outstanding treasury securities, the data for which is not publicly availableat a weekly frequency to the best of my knowledge. The use of the face-value data will onlyunderestimate the significance of my results. An increase is the supply of collateral will reduce theprice of the collateral. The measurement error in the increase in the supply of collateral will biasthe coefficient downward.






FIGURE 5 Spread between overnight treasury collateral repo rates and the ninety-daytreasury yield.

–1

0

1

2

3

2005W 26

2007W 1

2008W 27

2010W 1

2011W 26

2013W 1

Week

St =

rt –

yt3m

o

Figure 5 shows the spread between the overnight treasury collateral repo rates andthe three-month treasury yield. The fact that the spread was positive in the yearsleading up to the crisis, and has been close to zero since, seems to indicate thatcollateral has been less idle since the crisis.

3.4 Results

The identification assumption in the regressions is that the unobserved factors affect-ing the dependent variables are uncorrelated with the weekly changes in treasurysupply. The issuance of bonds by the treasury is according to a preset timetable,which does not interact with the unobserved factors at a weekly frequency.

The results for regression (3.3) are in Table 1. The spread between the repo rateand the risk-free rate increases significantly as the treasury supply increases, but notso much when the spread is already high and collateral is idle. Since the time seriesfor the spread indicates that collateral has been less idle since the crisis, I rerun theregression for before and after January 1, 2009. I find that the effect of treasury supplyon the spread is significant only after that date, when collateral was scarce. Figure 6shows this conditional correlation before and after that date.

The bilateral market is heavily characterized by rehypothecation. As shown inFigure 7, 75–80% of the collateral that is eligible to be rehypothecated is actuallyrehypothecated by the top five US broker-dealers. A change in the terms of borrowing





98 A. Muley

TABLE 1 Regressing repo-treasury spread on treasury supply.

�st=10�6

‚ …„ ƒBefore AfterJan 1 Jan 1

Whole period 2009 2009

�Tt 0.37* 1.06** 0.29 0.31**(0.15) (0.40) (0.65) (0.12)

st�1�Tt — �3.08*** — —— (0.65) — —

Controls Yes Yes Yes Yes

R2 0.28 0.32 0.31 0.37

N 383 383 179 204

Here, I use heteroscedasticity and autocorrelation consistent (HAC) standard errors.

FIGURE 6 Conditional correlation between treasury supply and the repo-treasury spread,before and after January 1, 2009.

2

1

0

–1

–2

0.2

0.1

0

–0.1

–0.2

Δst

Δst

–100 000 0 100 000 200 000ΔTt ΔTt

–100 000 0 100 000

(a) (b)

Conditioning on the control variables. (a) Before January 1, 2009. (b) After January 1, 2009.

collateral is, therefore, likely to have a significant effect on the ability of the dealersto satisfy their short-term liquidity needs through rehypothecation.

4 CONCLUSION

I construct a model of rehypothecation with the prime brokerage and repo in mind.I show that rehypothecation allows the intermediate lender to invest in their prof-itable project by effectively borrowing from the borrower by charging a high haircut.Funding chains in the prime-brokerage market typically exhibit segmentation, with






FIGURE 7 Collateral actually rehypothecated by top five US broker-dealers, as a fractionof collateral allowed to be rehypothecated.

0.5

0.6

0.7

0.8

0.9

1.0A

ctua

l/allo

wed

2007Q4

2008Q4

2009Q4

2010Q4

2011Q4

2012Q4

2013Q4

Quarter

the hedge fund borrowers being unable to directly borrow from the ultimate MMFlenders. I show that when the pledgeability of the intermediate lenders is limited,segmentation may achieve superior aggregate outcomes. The reason for this is apecuniary externality generated by the presence of other competing but unproductivelenders, who reduce the collateral available to the productive intermediate lenders toborrow and rehypothecate. I also show that open market operations of the central bankcan backfire, as high-quality collateral removed from the system can tighten collat-eral constraints down the rehypothecation chain and impair financing and investmentactivity. I discuss a number of implications for policy.

A direction for future research could be to further explore the quantitative effects ofopen market operations on collateral constraints and financing. Special repo spreadsprovide a more accurate measure of the discrepancy between the supply and demandfor collateral. Examining the impact of treasury supply on financing using treasurycollateral when the spreads are high will clarify the quantitative importance of theseeffects.







100 A. Muley

ACKNOWLEDGEMENTS

I would like to thank Alp Simsek, Robert Townsend and Ivan Werning for extremelyuseful discussions and comments. This paper has also benefited from discussionswith Manmohan Singh at the International Monetary Fund, and with Antoine Mar-tin, Tobias Adrian, Nina Boyarchenko, Nicola Cetorelli, Adam Copeland, FernandoDuarte, Thomas Eisenbach, Michael Fleming, James McAndrews, Asani Sarkar,James Vickery and other seminar participants at the New York Fed. The proofs ofthe results are included in an online appendix.

REFERENCES

Andolfatto, D., Martin, F. M., and Zhang, S. (2015). Rehypothecation and liquidity. WorkingPaper 2015-003, Federal Reserve Bank of St. Louis (http://doi.org/bmcw).

Araújo, A., Schommer, S., and Woodford, M. (2013). Conventional and unconven-tional monetary policy with endogenous collateral constraints. Working Paper W19711,National Bureau of Economic Research.

Bottazzi, J. M., Luque, J., and Páscoa, M. R. (2012). Securities market theory: posses-sion, repo and rehypothecation. Journal of Economic Theory 147(2), 477–500 (http://doi.org/fhxqjd).

Caballero, R. J. (2006). On the macroeconomics of asset shortages. Working PaperW12753, National Bureau of Economic Research (http://doi.org/fxh3td).

Caballero, R. J., and Farhi, E. (2013). A model of the safe asset mechanism (SAM):safety traps and economic policy.Working Paper W18737, National Bureau of EconomicResearch (http://doi.org/bmcv).

Caballero, R. J., Farhi, E., and Gourinchas, P. O. (2006). An equilibrium model of“global imbalances” and low interest rates. Working Paper W11996, National Bureauof Economic Research (http://doi.org/fzg9gn).

Carpenter, S., Demiralp, S., Ihrig, J., and Klee, E. (2015). Analyzing Federal Reserve assetpurchases: from whom does the Fed buy? Journal of Banking and Finance 52, 230–244(http://doi.org/bmct).

Cetorelli, N., Mandel, B. H., and Mollineaux, L. (2012).The evolution of banks and financialintermediation: framing the analysis.Federal Reserve Bank of NewYork Economic PolicyReview 18(2), 1–12.

Cochrane, J. H. (2014). Monetary policy with interest on reserves. Journal of EconomicDynamics and Control 49, 74–108 (http://doi.org/bmcs).

D’Amico, S., Fan, R., and Kitsul, Y. (2013). The scarcity value of treasury collateral: repomarket effects of security-specific supply and demand factors. Working Paper 2013-22,Federal Reserve Bank of Chicago (http://doi.org/bmcr).

Eren, E. (2014). Intermediary funding liquidity and rehypothecation as determinants of repohaircuts and interest rates. Working Paper, Stanford University (http://doi.org/bmcq).

Ewerhart, C., and Tapking, J. (2008). Repo markets, counterparty risk and the 2007/2008liquidity crisis. Working Paper 909, European Central Bank.

Fleming, M. J., Hrung, W. B., and Keane, F. M. (2010). Repo market effects of the termsecurities lending facility. Federal Reserve Bank of New York Staff Report 426, 1–25.






Goodfriend, M. (2002). Interest on reserves and monetary policy. Federal Reserve Bankof New York Economic Policy Review 8(1), 13–29.

Holmstrom, B., and Tirole, J. (1997). Financial intermediation, loanable funds, and the realsector. Quarterly Journal of Economics 112(3), 663–691 (http://doi.org/fqgxqc).

Infante, S. (2014). Money for nothing: the consequences of repo rehypothecation. WorkingPaper, Federal Reserve Board (http://doi.org/bmcp).

Kashyap, A. K., and Stein, J. C. (2012). The optimal conduct of monetary policy with inter-est on reserves. American Economic Journal: Macroeconomics 4(1), 266–282 (http://doi.org/fx9m87).

Keister, T., Martin, A., and McAndrews, J. (2008). Divorcing money from monetary pol-icy. Federal Reserve Bank of New York Economic Policy Review 14(2), 1–16 (http://doi.org/fxj7bb).

Krishnamurthy, A., and Vissing-Jorgensen, A. (2012). The aggregate demand for treasurydebt. Journal of Political Economy 120(2), 233–267 (http://doi.org/bkzg).

Krishnamurthy, A., and Vissing-Jorgensen, A. (2013). The ins and outs of LSAPs. In Eco-nomic Symposium Conference Proceedings. Federal Reserve Bank of Kansas City,Jackson Hole, WY.

Lee, J. (2015). Collateral circulation and repo spreads. Working Paper 2548209, SocialScience Research Network (http://doi.org/bmcn).

Martin, A., McAndrews, J., Palida, A., and Skeie, D. R. (2013). Federal Reserve tools formanaging rates and reserves. Federal Reserve Bank of New York Staff Report 642,1–34 (http://doi.org/bmcm).

Maurin, V. (2014). Re-using the collateral of others: a general equilibrium model ofrehypothecation. Working Paper, European University Institute.

Muley, A. (2016). Rehypothecation and monetary policy. Working Paper, MIT Economics.Pozsar, Z., Adrian, T., Ashcraft, A. B., and Boesky, H. (2010). Shadow banking. Federal

Reserve Bank of New York Staff Report 458, 1–35 (http://doi.org/fxnv6d).Singh, M. (2013). Collateral and monetary policy. Working Paper 13/186, International

Monetary Fund, pp. 1–17 (http://doi.org/bmck).Stein, J. C. (2012). Monetary policy as financial stability regulation. Quarterly Journal of

Economics 127(1), 57–95 (http://doi.org/fxjdm7).






Research Paper

Collateral chains and incentives

Charles M. Kahn1 and Hyejin Park2

1Department of Finance, University of Illinois, Urbana-Champaign, 340 Wohlers Hall,1206 South Sixth St, Champaign, IL 61820, USA; email: [email protected] of Economics, University of Illinois, Urbana-Champaign, 1407 West Gregory Drive,Urbana, IL 61801, USA; email: [email protected]

(Received May 17, 2016; revised June 9, 2016; accepted July 4, 2016)

ABSTRACT

Collateral reuse, either through explicit permission from the borrower or through arepo agreement, economizes on scarce liquid collateral but leaves the possibility ofmismatch of collateral allocation in the event of the failure of a party in the middle ofthe collateral chain. If haircuts are determined to solve incentive problems, there maybe a wedge between the shadow values of the collateral to parties in the collateralchain. This can tempt parties down the chain to overuse the collateral provided themand therefore cause parties up the chain to be unwilling to extend permission forreuse. We consider a variety of financial arrangements in light of this framework.

Keywords: collateral; rehypothecation; moral hazard; incentives; haircut.

1 INTRODUCTION

Most financial contracts are in the form of promises to pay a certain amount ofmoney or swap assets at a later date and at prearranged terms in exchange for someconsideration today. But often these promises cannot be warranted in themselves, and

Corresponding author: C. M. Kahn Print ISSN 2049-5404 j Online ISSN 2049-5412Copyright © 2016 Incisive Risk Information (IP) Limited

103




104 C. M. Kahn and H. Park

they need to be backed by an eligible asset or property – collateral1 – such as Treasurybills in repo transactions and residential houses in mortgage contracts.2 Collateralreduces the cost of finance to a borrower by providing protection to the lender. Butnot every asset is equally good for use as collateral. Other things being equal, the moreliquid and widely valued the asset the more acceptable it is as collateral.3 However,liquid assets are not in infinite supply, and in some situations they can be expensive –that is, there can be a significant premium on such assets. In particular, as the volumeof financial transactions has sharply increased over the last few decades, the demandfor collateral has also increased significantly, and economizing on the existing limitedamount of collateral has become an important issue for market participants.4

One important step is to allow the collateral to do “double-duty” – that is, to builda collateral chain, letting the lender reuse the collateral to finance a loan of their own.Instead of sitting idle in the lender’s account, tying up capital, the collateral effectivelybacks two loans at once. Or to use a monetary analogy: the velocity of transactionsis increased, allowing the same stock of financial assets to effect more payments inthe same amount of time (for an analysis along these lines, see Singh (2011)).

This phenomenon is pervasive and it arises in a variety of guises. In specificcases the institutional details vary considerably with the characteristics of borrow-ers, lenders and collateral assets. Sometimes the procedures are formally the reuseof pledged collateral; sometimes they are achieved through multiple sales and linked

1 It is important to note that throughout this paper we will be using the term “collateral” in itseconomic theory sense, which covers a much broader set of institutional arrangements than thestrict and specific legal definition. This is because our goal is to understand the unifying principlescovering these various institutional arrangements.2 Holmstrom (2015) notes that the oldest form of collateralized lending is the pawn shop:

The earliest documents on pawning date back to the Tang Dynasty in China (around650 AD)… The borrower brings to the pawn shop items against which a loan isextended. The pawn shop keeps the items in custody for a relatively short (negotiable)term, say one month, during which the borrower can get back the item in return forrepayment of the loan. It sounds simple, but it is a beautiful solution to a complexproblem.

For another insightful discussion on the origin of collateralized lending, see Geanakoplos (1996).3 On the other hand, for the borrower to have a desire to retain ultimate possession of the collateral,they must attach some special value to the collateral, even if the difference is only due to thedifferences in risk aversion or to a transaction cost in the asset market. Otherwise, the borrowerwould be better off raising funds by just selling the collateral.4 Krishnamurthy and Vissing-Jorgensen (2013) estimated the liquidity and safety premium on Trea-suries paid by the investors on average from 1926 to 2008 was 72 basis points per year, whichsupports the idea that there has been a large and persistent demand for safe and liquid assets in theeconomy. In a similar vein, Greenwood et al (2015) show that the monetary premium on short-termTreasury bills has a lower yield than would be predicted in a conventional asset-pricing literature.





Collateral chains and incentives 105

forward repurchases. The purpose of this paper is to lay bare common features ofthe phenomenon, to understand its costs and benefits, and to begin to understandthe extent to which agents’ decisions to build collateral chains do or do not promoteeconomic efficiency.

1.1 Why collateral

Collateral is central to the operation of the modern financial system. The circulationof collateral in the financial system occurs in a variety of ways, with legal distinctionsvarying across jurisdictions and practices varying across markets. One legal distinc-tion is whether the collateral is transferred as “pledged for reuse” or whether full titleis transferred between parties to the loan (see Singh 2014, Chapter 1). In the lattercase the lender becomes the owner of the asset, and their rights to it are unlimited andautomatic – this is the practice in many bilateral markets. In the former case, thereare a variety of rights that can be associated with collateral reuse, but the terms ofreuse must be specifically laid out in the pledge agreement. “Rehypothecation” refersspecifically to the repledging of collateral, as opposed to broader categories of reuse,such as selling or lending.

Singh (2010, 2011) emphasizes the role of collateral in the financial market anddiscusses how reuse of collateral changed after the financial crisis. For example, beforethe crisis, he estimates the value of the total amount of rehypothecatable asset heldby the largest US investment banks, was estimated around US$4.5 trillion, droppingto US$2 trillion by 2009.

Collateral is central to recent changes in regulation of financial markets. For exam-ple, in most swap contracts the Dodd–Frank Act requires collateral to be held in asegregated account of a central counterparty (CCP). Nevertheless, there are funda-mental differences in the treatment of collateral and its reuse across countries. Forexample where the rights to reuse in Europe (including the UK) are strongly protectedand typically included in arrangements with broker-dealers, rights to rehypothecationare strictly limited in the United States. In the United States, SEC rule 15c3-3 allowsa prime broker to rehypothecate assets up to 140% of the value of the clients’ liability,but in the UK, there is no limit on the amount that can be rehypothecated.

On the other hand, it is important to remember that there is always an alternativeto collateralized lending: if the borrower owns collateral, they can sell it and use theproceeds to whatever end they had in mind. If they desire to reobtain the collateralthey can repurchase it in the future or (if there is a forward market available) makethe repurchase at the same time. Thus the likelihood of collateralized lending takingplace at all depends on the benefits it bestows.

As emphasized by Mills and Reed (2012), collateral in a lending relationship servestwo fundamental roles, as hostage and as insurance. First, collateral provides a bor-rower with incentives to repay to avoid forfeiting it; second, collateral provides a






lender with recourse to collect some revenue by liquidating it in the event that theborrower defaults. The value as insurance depends on the value of the collateral to thelender; in particular on the ability to resell the collateral easily. The importance of thecollateral as insurance depends on the likelihood that the borrower will be unable torepay and the risk aversion of the lender; if this likelihood is remote then the value tothe lender is immaterial. On the other hand, the importance of the collateral as hostagedepends on the likelihood that the borrower will be tempted to choose not to repay,or more generally, choose to take steps which render them unable to repay. The valueas hostage depends on the value of the collateral to the borrower. If the collateral isirreplaceable, the borrower will take care not to jeopardize its return.5

1.2 Plan of the paper

In this paper we build an extremely basic model of a three-agent collateral chain: Aenters into a collateralized debt arrangement borrowing from B, who subsequentlyenters a collateralized debt arrangement with C. In the model collateral plays a roleboth as hostage and as insurance. However, in order to keep the results as simple andstark as possible, we focus on a single role for collateral in each part of the chain:collateral will act as a hostage in the relation between A and B, and it will act asinsurance in the relation between B and C.6

Depending on parameter values, allowing B to build a collateral chain (for example,by rehypothecating the collateral) may improve efficiency in the economy, by allowingdesirable investments to take place. However, for some parameter values, the benefitsthat B obtains from building the collateral chain are outweighed by the damage thisdoes to A. Thinking in terms of rehypothecation, if A and B cannot write a contractthat adequately limits the conditions under which rehypothecation occurs, then Amay prefer to prohibit rehypothecation entirely. The model provides conditions underwhich this will be the case.

The structure of the paper is as follows. In Section 2 we provide a brief overview ofseveral strands of the literature to which this paper relates. Section 3 starts the model:it provides a simple structure in which using collateral as a hostage allows A to make aproductive investment, and no alternative to collateralized lending – and, in particular,neither uncollateralized lending nor outright sale of collateral – will work. In order tomaintain the borrower’s incentives, the collateralized loan will include a “ haircut”: the

5 The same two considerations apply on the lender’s side as well, although not in parallel: howsafely protected is the collateral from failure by the lender or a strategic choice by the lender not toreturn it? On the one hand, if the lender places little value on the collateral, they will be likely tobe willing to return it, and if the borrower values it little, they will not be so bothered if the lenderturns out to be unable to return it.6 However, our results continue to hold if we assume a more general structure.






amount borrowed will be less than the value of the collateral. Section 4 adds in the thirdagent and considers the incentives for B and C to establish a collateralized loan whenB’s activities are risky and the collateral serves as insurance against failure to repay.We contrast the optimal contract between B and C taking B’s previous contract withA as given, with the overall efficient outcome, taking into account A’s potential losses.

In the model, uncollateralized lending is infeasible and the structure is so sim-ple that the competing alternative to collateralized lending is a simple spot sale andrepurchase of the asset. Of course our model abstracts from many important aspectsof collateralized lending and in more complex environments there are additional alter-natives to consider. Section 5 examines a variety of alternatives in order to pinpointthe key feature of collateralized lending, showing how our basic result should bereinterpreted in a richer environment. Section 6 focuses on the institutions that canbe used to remedy the inefficiencies that we have shown to arise in collateral chains.Section 7 concludes.

2 LITERATURE REVIEW

This paper is closely related to the literature that analyzes how collateral can help toprovide funding liquidity when the borrower is subject to the limited pledgeabilityproblem: see, for example, Mills and Reed (2012), Biais et al (2016) and Boltonand Oehmke (2015). In all three of these papers, collateral mitigates the borrower’smoral hazard problem by strengthening the incentive constraint.A key feature in thesemodels is that collateral is transferred from the borrower to the lender at the time thecontract begins and sits in the lender’s account until the borrower buys it back at alater date – in other words, collateral in these models is similar to a repo agreement.We contribute to this literature by considering the additional risk that the lender mightfail to return collateral to the borrower (which occurs naturally from the process ofreusing collateral in our model), and by showing how this mutual lack of commitmentaffects the terms of contract between these parties.

This paper also relates to the recent literature that studies the endogeneity of col-lateralized borrowing in the environment where asset sale is possible, for example,Monnet and Narajabad (2012) and Parlatore Siritto (2014). Monnet and Narajabad(2012) show that agents prefer using repo agreements to sell the asset if they faceuncertainty about the future value of the asset. Parlatore Siritto (2014) provides theenvironment where even when collateralizable asset is liquid (in the sense that everyagent values the asset the same in autarky), in equilibrium, the borrower values theasset higher than does the lender and chooses using collateralized debt rather thanselling the asset.

Our paper builds on the idea that assets acceptable as collateral are scarce. For exam-ple, Krishnamurthy and Vissing-Jorgensen (2013) show that there exists a persistentpremium on Treasury bonds beyond the level that can be rationalized by a standard






asset pricing model. Gorton and Ordonez (2013) assert that there is a demand for agovernment-provided safe asset that can be used as collateral and the economy canbe fragile if their supply is inadequate to meet private demand.

Finally, our paper is a part of the literature analyzing rehypothecation, most ofwhich emerged after the financial crisis. Monnet (2011) discusses economic benefitsand costs of rehypothecation and regulatory concerns regarding it. Bottazi et al (2012)show how the leverage can increase in the process of rehypothecation.Andolfatto et al(2015) argue that restriction on rehypothecation is generally socially optimal. Maurin(2014) shows that rehypothecation is effective only when the market is incomplete.Lee (2015) shows how a sudden decline of rehypothecation rate might lead the systemto be fragile against a small shock. On the other hand, Eren (2014) and Infante(2014) show how the intermediary generates liquidity by setting a different marginon collateral providers and cash lenders. Muley (2016) provides conditions underwhich rehypothecation is more likely relative to securitization.

3 A SIMPLE MODEL OF COLLATERALIZED LENDING 7

In this section we develop a simple structure in which, under some conditions, usingcollateral as a hostage allows a borrower to make a productive investment, whenneither uncollateralized lending nor outright sale of collateral will work.

Consider two individuals: A and B, both risk-neutral with no discounting. A has aproject they wish to finance today. The project provides a payoff of R > 1 per unit ofinvestment. A needs to raise funding; B has funds. However, B has no direct way toforce A to repay what they borrow from B, so uncollateralized lending is infeasible.

Suppose A possesses one unit of an asset which has a market value8 of Z0 andagents bear a transaction cost of TM in purchasing or selling the asset on the market.9

7 This is a simplified version of Kahn and Park (2015).8 In our simple model the future market price of the asset is certain, and with risk neutrality and nodiscounting this means the future price is the same as the market price today. While this is a drasticsimplification it does not have an important effect on the model. If there is uncertainty about thefuture market value of the collateral then each of the parties may want to enter forward contractsto insure themselves, but with forward prices we again reduce the problem to one where futureasset price uncertainty is eliminated. More complex considerations arise when incomplete forwardmarkets lead to an incentive to walk away from the contract if the price of the collateral collapses. Inthis case real world contracts provide for more complex insurance arrangements between the partiesthemselves. However, the key to the use of collateral as hostage is not the insurance component butthe difference between the expected value of the collateral and the value of the loan. These issuesare discussed in greater detail in Section 5.9 In other words, the collateral in our model is less than perfectly liquid, although this transactioncost can be small. Transactions costs can arise indirectly from many sources, including portfolioconsiderations, costs of resale, and uncertainty of title.






We assume that the asset has a private value to A of Z > Z0 C TM (thus A wishes toown the unit of the asset at the end of the period) but it is worth less than Z0 � TM toB (thus B does not wish to retain it at the end of the period).10

Given that they want to hold the asset in the future, A has two options to raisefunding for their investment. One possibility is to sell the asset now, and repurchaseit after the investment is completed. Then A’s payoff from the arrangement is

R.Z0 � TM/ C Z � .Z0 C TM/

(and this is better than selling and not repurchasing provided Z > Z0 C TM).An alternative is a collateralized loan: provided B is completely trustworthy, A

can deliver their asset temporarily to B, in exchange for funding for A’s project, andthen repay when the project matures and get their collateral back. Since B can collectZ0 � TM by selling it on the open market in the event that the borrower defaults, itwould be perfectly safe for B to lend that amount to A in return. We will assume thelending environment is competitive (in other words, B will receive zero profits in anybilateral contract with A). Then A’s payoff is

.R � 1/.Z0 � TM/ C Z � TL;

where TL is the transaction cost associated with the loan (assume TL is paid by A).Note, therefore, that one immediate source of advantage of collateralized lendingcould simply be the savings in transaction costs, which is represented by the spreadof 2TM � TL. But even if this spread were zero, there would still be a slight advantagein the collateralized loan. Recall that the collateral can be used as hostage, and itsvalue depends on the value to the borrower. In the simple case at hand, this meansthat A can credibly borrow the replacement cost of their collateral, which is slightlymore than the market value.

However, in practice, it is common for the transaction to include a “haircut”: thevalue of the loan is less than the value of the collateral backing it. An importantrationale for haircuts is moral hazard: the difference deters undesirable risk taking.11

For simplicity we will use the most basic version of a moral hazard problem as anillustration; for a more standard, but more complicated, approach, see Kahn and Park(2015).

10 The collateral has to be worth more to A than to B; otherwise A would simply sell it outright toB. There are many possible sources of difference between the valuations. For example, ParlatoreSiritto (2014) develops an environment where differences in valuations arise because the asset canbe used by some agents for future investment opportunities, even though the asset gives the samecurrent dividends to everybody.11 A second reason for haircuts (or margin) is protection against fluctuation in the value of thecollateral. We will turn to that consideration in the discussion below in Section 5.






Suppose the borrower has the option of diverting the funds they receive to theirown benefit (“absconding” as in Calomiris and Kahn (1991)) and the benefit receivedthereby is a per unit of funds diverted, which is also assumed to be non-tradable (ineffect A cannot repurchase the asset if they divert the funds). Appendix A, availableonline, provides conditions under which the following hold.

� The optimal contract between A and B is a collateralized debt contract (I1; X1)under which A receives an amount I1 initially in return for temporarily hand-ing over the collateral and receives the collateral back when they pay B theprearranged amount X1 at the maturity of their investment.

� The loan is “overcollateralized” in the sense that the market value of thecollateral exceeds Z0 exceeds the loan amount I1.

� A prefers the collateralized loan to selling the collateral outright.

The size of the haircut in this optimal contract depends on the temptation providedby funds diversion: the greater the temptation the stricter the haircut (and so thesmaller the loan that is feasible for a fixed amount of collateral). Thus the optimalityof collateralized lending depends on the relative size of the transaction costs and theextent to which the incentive problem imposes a haircut on the loan.

We next consider the implications if B faces a risk of failure.12 For concretenesswe imagine that with probability 1 � � , B “damages” the collateral and is unable toreturn it to A. In this case A does not repay the loan. We assume that the salvagevalue of the collateral to B is Zs 6 Z0. The possibility that the collateral is damagedand stranded with B reduces the desirability of collateralized lending. Because theexpected value from the return of the collateral is lower, the haircut needed to keep theborrower honest must be correspondingly greater – that is, the maximum feasible loanfor a given amount of collateral now shrinks. Appendix B, available online, providesthe stricter set of conditions under which collateralized lending remains the optimalchoice.

However, as long as A correctly anticipates the risk of failure by B, the outcomewill be efficient: A and B reach a contract if and only if collateralized lending is theconstrained optimal action.

4 COLLATERAL CHAINS

Now suppose that B has a project as well and, after having provided financing forA, finds themselves in need of financing in turn. At that point their sole resource

12 Recall that we are not including an insurance motive for B’s holding A’s collateral – that is, thereis no risk associated with A’s project.






is the collateral received from A. Whether they are allowed to reuse the collateraldepends on the terms agreed between them and A. If it is not possible to reuse thecollateral, that is the end of the story. In this section we will consider the arrangementB makes for financing from a third party, C, under the assumption that collateral reuseis permitted, and then consider how foreknowledge of this activity affects A and B’sinitial contractual decision.

For the collateralized loan between B and C, there are again in general the twinconcerns of insurance and incentive. While we concentrated solely on the incentiveaspect in dealing with A’s loan, we will focus solely on the insurance aspect of B’sloan. We assume that B’s project costs 1, has a gross return of Y if successful andof 0 if unsuccessful, and the probability of success is � . Finally we assume that thecollateral is worth Z0 to C, and for simplicity assume neither B nor C is subject totransaction costs in dealing with each other. The conditions of the contract are asbefore: B receives a unit of financing from C to invest in their project; if it succeeds,they pay the prearranged amount to C in return for the collateral. If the project fails,C keeps the collateral. With competitive loan markets the payment to C if the projectsucceeds is

X2 D ��1.1 � .1 � �/Z0/:

B’s ex post profits13 from the decision to take the loan are

�.Y C X1 � X2/;

while their profits if they do not take the loan are X1. Thus, they will decide to takethe loan if

�Y > .1 � �/X1 C �X2

D 1 � .1 � �/.Z0 � X1/: (4.1)

As we have seen, the investment choice ofA when contracting with B is constrainedefficient. In contrast let us consider the efficiency of the investment decision of B.

Since B has no source of funding except by using A’s collateral, there are twocomponents to B’s costs, the cost of the funds themselves and the possible damage tothe collateral. Here the “damage” comes from the fact that the failure of B’s projectwill leave the collateral in the hands of C, requiring A to make an expenditure toreplace it. Thus the social opportunity cost of the investment is

1 C .1 � �/TM;

and the project ought to be funded if and only if

�Y > 1 C .1 � �/TM: (4.2)

13 At this point I1 is a sunk cost.






Now, from the two conditions and (4.1) and (4.2) above, we see that if the haircut onthe initial loan is positive, it is possible for B to choose to take on socially unprofitableprojects. In particular, this will occur if and only if

1 C .1 � �/TM > �Y > 1 � .1 � �/.Z0 � X1/:

Comparing the left and the right side of this equation shows that whether B is likelyto be interested in undertaking inefficient projects depends on the relative size ofTM, X1 and Z0. In particular, when either A’s cost of replacing the collateral, TM,or the degree of overcollateralization Z0 � X1, increases, then so does the range ofinvestments over which B chooses an inefficient decision regarding their reuse of thecollateral.

The inefficiency in the contract between B and C arises from the externality that B’sfailure imposes on A. If the risk of failure is known at the time that A and B first meetthenA can accept or reject the contract at that point. But if B’s investment is inefficient,it would in general be better still for A to accept collateralized loan while refusingto grant permission to B to pass the collateral forward. More generally, if the risk �

is not known at the time that A and B write their contract, A might wish to establishminimum standards under which B is permitted to rehypothecate A’s collateral. Withovercollateralization, these standards will be stricter than those B would choose touse in making their decision ex post. If establishing such standards is infeasible, thendepending on the distribution of � , A may prefer to prohibit B from rehypothecatingthe collateral.

5 DISCUSSION

In the model, uncollateralized lending is infeasible and the competing alternative tocollateralized lending is a simple spot sale and subsequent repurchase of the asset.Of course the model abstracts from many important aspects of collateralized lendingand in more complex environments there are additional alternatives to consider. Thusfor comparison it is useful to consider institutional arrangements other than spot saleand subsequent spot repurchase.

(1) The next simplest possibility is spot sale and independent forward purchaseof the collateral. Agent A insures the terms at which repurchase will occur byestablishing the forward contract at the same time as the spot sale. However,the transactions are made with unrelated parties, so no cross subsidization ispossible.

(2) A more complex arrangement would be to link the spot sale and forward con-tract. This arrangement is equivalent to our notion of a collateralized loan. Once






the contracts are linked (as in a repo transaction), the two portions need notindividually break even. For example, when there is a haircut in the arrange-ment, the unequal trade at the end, whereby the borrower gets back somethingmore valuable than their repayment, subsidizes the initial unequal deal in theopposite direction. It is this cross-subsidization through overcollateralizationthat is key to the effects of a collateral chain, because this is the feature whichboth provides incentives to the borrower and leads to conflict of interest whenthe lender tries to reuse the collateral.

5.1 Additional comments

Once the two contracts are linked, a further important distinction arises as to thecontingencies under which the second portion of the contract can fail.

(3a) Collateral remote from lender. The collateral is held in such a way that even ifthe lender itself fails, the collateral returns to the borrower when the borrowerfulfills their conditions.

(3b) Collateral integrated into the lender. The collateral is held in such a way thatretrieval becomes expensive or impossible if the lender fails. In effect the bor-rower substitutes the counterparty risk for the market risk in retrieving thepledged asset.14

In practice a variety of techniques are available for making the collateral inte-grated or remote. For us the fundamental feature of collateral is that it represents anobligation incurred by the lender to return title to the borrower of certain previouslyacquired assets if repayment conditions are met. At one extreme, these assets couldbe held by a trustee segregated from the rest of the lender’s business. At the otherextreme, initial title could be transferred outright to the lender (including full rights toresell). Legal structures of various sorts in various jurisdictions (repos, for example,

14 This distinction also arises on the flip side: the extent to which the particular legal arrangementemployed permits or restricts the ability of the lender to benefit by liquidating the collateral in theevent of failure by the borrower. The right to liquidate pledged collateral in a secured loan mustbe specifically agreed on; in a repo agreement the ability to liquidate is generally greater, althoughsubtle differences exist between European and US rules. There is always a danger that in bankruptcythe arrangement could be recharacterized as a secured loan, subject to restrictions on liquidation(see, for example, International Capital Market Association (2013) for more details). A related focusof the law and finance literature is on the advantages and disadvantages of various “safe-harbor”provisions in bankruptcy, and the extent to which they should be overriden in the case of financialassets. Note that this is an issue that is important to the extent that collateral is used in its insurancerole; it does not pertain when collateral is being used strictly for incentives; there the issue is notthe ability of the lender to enjoy it but the ability to prevent the borrower from enjoying it. (For anexamination of the implications of these issues for financial stability, see Antinolfi et al (2015).)






or as emphasized by Muley (2016) special purpose vehicles for mortgages) providea variety of intermediate degrees of remoteness.

For real assets with extreme idiosyncratic characteristics, the borrower will wantthe return of the specific asset entrusted to the lender – it is not acceptable for a pawnshop, for example to return to the borrower a wedding ring “just as good as” theone pledged. For assets with less inherent specificity – and in particular for financialassets, the promised delivery is not of the specific unit pledged but merely of anindistinguishable unit. The more likely the lender is to fail and therefore not returnthe collateral, the less likely collateralized lending will dominate sale.

In the extreme, in many short-term financial market arrangements, the importantfeature of the pledged asset is simply its suitability for use in recurring transactions.In this case the promise of the return of a “ similar” liquid asset will suffice (as intri-party repos).

Our model has ignored the second important reason for the existence of haircuts:the risk of changes in the value of the collateral to the borrower. If the borrower’svaluation of the collateral falls too much, they may prefer ex post to walk away fromthe obligation, reducing the usefulness of collateralized lending. The haircut makesit less likely that the borrower will choose to walk. While this additional incentiveproblem complicates the search for the optimal contract, it does not alter our basicstory – it simply adds a second component to the haircut. In the case of many real assets(for example, owner-occupied housing) the moral hazard problem – maintenance andupkeep of the asset – is probably a dominant consideration, since the differencebetween the borrower’s valuation of the asset and the market valuation is likely tobe large. However, in the case of liquid financial assets, the premium placed on thepledged asset by the borrower will be small, and so fluctuations in market prices ofcollateral can affect the willingness to repay the collateralized loan. The solution usedin financial markets is marking the contract to market with sufficient frequency. In suchan environment the measure of the haircut associated with incentives in collateralizedlending as we have described it would in principle be the difference between thehaircut in the relationship contract and the margin in a standalone arms-length futurescontract. This differential could, of course, go either way depending on the incentivesand reliability of the two parties.

6 REMEDYING INEFFICIENCY IN COLLATERAL CHAINS

On the one hand, the ability to construct collateral chains is valuable. The prospectof being able to reuse the collateral increases the value to B of the initial loan, andso will decrease the expense of the loan to A – in this case increasing the amountthat A can borrow. On the other hand, by passing some rights along to a third party,the collateral chain increases the cost or difficulty of A’s reacquiring the collateral,






should B fail. Whether collateral chains are efficient depends on the costs and benefitsof these two considerations.

When liquidity is short, reuse becomes more important. Whether B is likely tomake an efficient decision about their reuse of the collateral depends on how closelythe haircut on the collateral matches A’s cost of replacing the collateral. If the haircutis much larger than replacement cost, then A will want to preclude B’s ability toreuse collateral, and agreements between A and B will restrict B’s flexibility. As thepledgeable asset becomes more liquid – that is, as the transaction costs for its purchaseor sale become lower – the effects on collateral chains are ambiguous. On the onehand, the reduced costs of misallocation make it less costly to use collateral chains;on the other hand, the liquidity of the asset allows the initial borrower to raise fundsby selling it directly rather than by using it for collateral.

When misallocation is expensive, financial markets have incentives for developingtechniques to make the unwinding of complicated chains less onerous. Kahn andRoberds (2007) examine credit chains in the context of transferable debt. In theiraccount, a credit chain is also established, but then the middleman in the chain isreleased, passing the IOU received from their debtor to their creditor. The use oftransferable debt allows the avoidance of a multistep transfer of the settlement assetand the reaping of some efficiency and incentive gains. While the two accounts differin one important way – debts in Kahn and Roberds are uncollateralized – a naturalquestion nonetheless arises: could a similar improvement be realized in the presentenvironment, by eliminating B from the collateral chain and allowingA to deal directlywith C? This is all the more tempting because the inability of A and C to deal directlyis the source of the inefficiencies in B’s decisions, and B’s location between themseems to be a source of monopoly power as well.

The answer to this question depends on, among other things, the structure of infor-mation in the economy. To the extent that no incentive problem exists, there is no lossin pulling B out of the collateral chain and allowing A and C to deal directly. However,if the source of B’s centrality is information aboutA’s and C’s creditworthiness, then itmay not be possible to drop B from the chain while retaining the useable information.

On the other hand, to the extent that the primary risks are associated, not with theincentives of the parties, but with the collateral itself, an arrangement that substitutes asafer holder of collateral for the risky party B would unambiguously improve the work-ings of the economy. This is what a central counterparty accomplishes: substitutingthe (presumably safer) specialized or mutualized intermediary for the counterpartiesin an exchange (for an examination of some of the incentive problems that can arisein such an environment, see Koeppl (2013)).

Finally, note again that these considerations do not depend on the formal legalstructure of the collateral chain; instead they depend on the ability of B to pass rights to






the collateral on to a third party – either by repledging or reselling it, while maintainingthe obligation to provide the asset or its equivalent to the initial borrower.15

7 SUMMARY

Collateral is useful as incentive and insurance, but good collateral can be in limitedsupply and therefore expensive; thus economies develop ways to economize on itsuse. Reuse is one method of increasing the velocity of liquid collateral. A borrowercan give explicit permission for reuse of collateral or, equivalently, they can arrangea coordinated sale of the collateral and repurchase (of it or of equivalent assets) fromthe same counterparty.

Haircuts in collateralized lending are important for guaranteeing that the value ofthe collateral does not stray too far below the promised repayment. But haircuts alsoserve as a disciplinary device when parties to the contract are subject to moral hazard.Because of the haircut, there can be a wedge between the replacement cost to theborrower and the cost of reusing or selling the collateral as perceived by the lender.If this wedge is significant, the borrower will want to withhold the lender’s ability toreuse the collateral, preferring arrangements in which the asset is immobilized.

The costs involved in collateral reuse arise from the possibility of misallocation ofthe collateral in the event of failure of someone intermediate in the collateral chain. Ifthese possibilities are remote, or if the cost of the misallocation is small because of theease with which the collateral can be replaced, then techniques for reusing collateralare likely to be socially beneficial.



ACKNOWLEDGEMENTS

We thank participants at the Atlanta Federal Reserve Conference in Amelia Island,Florida, for valuable suggestions, and in particular we thank Phil Prince andManmohan Singh, as well as the anonymous referee.

15 In this respect the problem has important similarities to the phenomenon of deposit creation bybanks as described, for example, in Parlour et al (2016): the bank takes the funding it has receivedfrom depositors and uses them to back, at the stroke of a pen, the liquidity of new borrowers as well.






REFERENCES

Andolfatto, D., Martin, F. M., and Zhang, S. (2015). Rehypothecation and liquidity. WorkingPaper, Federal Reserve Bank of St. Louis (http://doi.org/bmcw).

Antinolfi, G., Carapella, F., Kahn, C., Martin, A., Mills, D., and Nosal, E. (2015). Repos, firesales and bankruptcy policy. Review of Economic Dynamics 18(1), 21–31 (http://doi.org/bm6z).

Biais, B., Heider, F., and Hoerova, M. (2016). Risk-sharing or risk-taking? Counterpartyrisk, incentives and margins. Journal of Finance 71(4), 1669–1698 (http://doi.org/bm62).

Bolton, P., and Oehmke, M. (2015). Should derivatives be privileged in bankruptcy? Journalof Finance 70(6), 2353–2394 (http://doi.org/bm63).

Bottazzi, J.-M., Luque, J., and Pascoa, M. R. (2012). Securities market theory: possession,repo and rehypothecation. Journal of Economic Theory 147(2), 477–500 (http://doi.org/fhxqjd).

Calomiris, C., and Kahn, C. (1991). The role of demandable debt in structuring optimalbanking arrangements. American Economic Review 81, 497–513.

Eren, E. (2014). Intermediary funding liquidity and rehypothecation as determinants of repohaircuts and interest rates. Paper, 27th Australasian Finance and Banking Conference(http://doi.org/bmcq).

Geanakoplos, J. (1996). Promises promises. Discussion Paper 1143, Cowles Foundationfor Research in Economics, Yale University.

Gorton, G. B., and Ordonez, G. (2013). The supply and demand for safe assets. TechnicalReport, National Bureau of Economic Research (http://doi.org/bm64).

Greenwood, R. M., Hanson, S. G., and Stein, J. C. (2015). A comparative-advantageapproach to government debt maturity. Journal of Finance 70, 1683–1722 (http://doi.org/bm65).

Holmstrom, B. (2015). Understanding the role of debt in the financial system. WorkingPaper, Bank for International Settlements. URL: http://ssrn.com/abstract=2552018.

Infante, S. (2014). Money for nothing: the consequences of repo rehypothecation. WorkingPaper (http://doi.org/bmcp).

International Capital Market Association (2013). Frequently asked questions on repo, 19:why is it important to document repo? URL: www.icmagroup.org.

Kahn, C., and Park, H. (2015), Collateral, rehypothecation, and efficiency. Working Paper,University of Illinois.

Kahn, C., and Roberds, W. (2007), Transferability, finality and debt settlement. Journal ofMonetary Economics 54(4), 955–978 (http://doi.org/fwj3wk).

Koeppl, T. (2013). The limits of central counterparty clearing: collusive moral hazard andmarket liquidity. Working Paper, Queen’s Economics Department.

Krishnamurthy, A., and Vissing-Jorgensen, A. (2013). Short-term debt and financial crises:what we can learn from US treasury supply.Unpublished paper, Northwestern University.

Lee, J. (2015). Collateral circulation and repo spreads. URL: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2548209

Maurin, V. (2014). Re-using the collateral of others: a general equilibrium model ofrehypothecation. Working Paper, European University Institute.

Mills, D. C., Jr., and Reed, R. R. (2012). Default risk and collateral in the absence ofcommitment. Unpublished manuscript (http://doi.org/fxt3xg).






Monnet, C. (2011). Rehypothecation. Business Review, Federal Reserve Bank of Philadel-phia (Q4), 18–25.

Monnet, C., and Narajabad, B. (2012).Why rent when you can buy? A theory of repurchaseagreements.2012 Society for Economic Dynamics Meeting Papers, Number 647.

Muley, A. (2016). Collateral reuse in shadow banking and monetary policy. Working Paper,Massachusetts Institute of Technology.

Parlatore Siritto, C. (2014). Equilibrium collateral constraints. Working Paper, URL: http://bit.ly/2axvw5y.

Parlour, C., Rajan, U., and Walden, J., (2016). Making money: commercial banks, moneycreation and the payment system. Working Paper, Haas School of Business, Universityof California at Berkeley.

Singh, M. (2010). The sizable role of rehypothecation in the shadow banking system.Working Paper, IMF (http://doi.org/bm66).

Singh, M. (2011). Velocity of pledged collateral: analysis and implications. Working Paper,IMF (http://doi.org/bm67).

Singh, M. (2014). Collateral and Financial Plumbing. Risk Books, London.





Documents

Financial Market Infrastructures - Risksubscriptions.risk.net/.../JournalofFinancialMarketInfrastructures.pdf · The Journal of Financial Market Infrastructures EDITORIAL BOARD Editor-in-Chief