Multiplex network analysis of the UK over‐the‐counter derivatives market

• Author: Gerardo Ferrara,Marco Bardoscia,Ginestra Bianconi
• DOI: http://doi.org/10.1002/ijfe.1745
• Published date: 01 October 2019
• Date: 01 October 2019

RESEARCH ARTICLE

DOI: 10.1002/ijfe.1745

Multiplex network analysis of the UK over‐the‐counter

derivatives market*

Marco Bardoscia

1

| Ginestra Bianconi

2,3

| Gerardo Ferrara

1

1

Bank of England, London, UK

2

School of Mathematical Sciences, Queen

Mary University of London, London, UK

3

The Alan Turing Institute, British

Library, London, UK

Correspondence

Gerardo Ferrara, Bank of England,

London, UK.

Email: Gerardo.Ferrara@bankofengland.

co.uk

*The viewsexpressed in this paper are those

of the authors and notnecessarily those of

the Bank of England,Queen Mary

Universityof London,Alan Turing Institute,

or any other institution with whichthe

authors may be affiliated or associated.

Andrew Haldane,Mark Manning, Sean

McGrath, AngusMoir, Pedro Gurrola‐

Perez, RadoslavRaykov, Andrea Serafino,

Yedidiah Solowiejczyk, John Tanner,

Michalis Vasios,Nicholas Vause, Paul

Nahai‐Williamson,Michael Yoganayagam,

and an anonymousreferee provided

valuable comments. We are particularly

grateful forthe assistance given by George

Bartonand Katia Pascarellain collecting the

data. All errorsremain those of the authors.

Abstract

In this paper, we analyse the network of exposures constructed by using the UK

trade repository data for three different categories of contracts: interest rate,

credit, and foreign exchange derivatives. We study how liquidity shocks related

to variation margins propagate across the network and translate into payment

deficiencies across different derivative markets. A key finding of the paper is

that, in extreme theoretical scenarios where liquidity buffers are small, a handful

of institutions may experience significant spillover effects due to the directional-

ity of their portfolios. Additionally, we show that two novel multiplex centrality

measures, the Functional Multiplex Eigenvector Centrality and the Functional

Multiplex PageRank, can be used as a proxy for the vulnerability of financial

institutions, with the Functional Multiplex PageRank improving on the results

that can be obtained using the Functional Multiplex Eigenvector Centrality.

KEYWORDS

clearing house (CCP), financial networks, liquidity shock, multiplex networks, systemic risk

JEL CLASSIFICATION

D85; G01; G17; L14

1|INTRODUCTION

Since the global financial crisis in 2007, the G20

1

has over-

seen an ambitious program of regulatory reform in finan-

cial markets. One goal of the reform program is to make

derivative markets safer by reducing systemic risk and

improving counterparty risk management. For this reason,

many standardized over‐the‐counter (OTC) derivative con-

tracts must now be cleared through clearing houses

(CCPs). For example, in the EU, specific classes of interest

rate derivatives

2

must be cleared through CCPs. More

recently, as CCPs have increased notably in their size and

many institutions have become more exposed to CCPs,

authorities are examining the role CCPs may play as a

source of stress in the financial system (Bank for Interna-

tional Settlements, 2018; Alfranseder et al., 2018).

In this paper, we analyse the nature of the

interconnectedness in the UK derivative markets to

understand existing interdependencies across different

derivative markets and to give a high‐level overview of

potential channels for the transmission of liquidity shocks

in the system. In order to do this, we follow two

------------------------------------------------------- -- --- -- -- --- -- --- -- -- --- -- --- -- -- --- -- --- -- -- --- -- --- -- --- -- -- --- -- --- -- -

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the

original work is properly cited.

© 2019 Bank of England. International Journal of Finance & Economics © 2019 John Wiley & Sons Ltd.

1520 Int J Fin Econ. 2019;24:1520–1544.wileyonlinelibrary.com/journal/ijfe

BARDOSCIA ET AL.1521

complementary approaches. First, we represent the UK

derivative markets as a multiplex network

3

(Bianconi,

2018) and we use state‐of‐the‐art multiplex centrality mea-

sures to assess the vulnerability of financial institutions.

Second, we draw on the recent contributions on contagion

mechanisms in OTC derivative markets to gain a better

understanding of the transmission of liquidity stress.

This paper extends the existing academic literature on

the OTC derivative markets in three main directions. First,

we use UK trade state reports fromthe Depository Trust &

Clearing Corporation (DTCC)and Unavista trade reposito-

ries (as of 30th June 2016) that include data on both cen-

trally and noncentrally cleared trades of CCP clearing

members. This allows us to build a network of exposures

between those institutionsacross three OTC markets: inter-

est rate, credit, and foreign exchange derivatives. To the

best of our knowledge, Abad et al. (2016) is the only other

study on this same set of asset classes, but it only offers a

comprehensivesurvey of each derivative market separately.

In contrast, we analysethe structural properties of the three

derivative markets simultaneously in a multiplex form.

Second, in order to quantitatively estimate the vulner-

ability of financial institutions we develop and compare

two extensions of the eigenvector centrality and of the

PageRank centrality. These extensions are the Functional

Multiplex Eigenvector Centrality (FMEC) and the Func-

tional Multiplex PageRank (FMP), which are based on

the work of Iacovacci et al. (2016).

Third, we extend the analysis first developed by Paddrik

et al. (2016) from a financial network with only one layer

for the credit default swaps (CDSs) to a multiplex network

made of three different markets–interest rate, credit, and

foreign exchange derivatives, thereby capturing the inter-

actions between them. We computethe impact of variation

margin (VM) shocks in terms of potential deficiencies in

expected payments between market participants.

We observe that the FMP centrality ranking offers a

small improvement identifying the most vulnerable insti-

tutions involved in the propagation of VM shocks in the

system, compared with the FMEC centrality. Additionally,

our contagion model allows us to estimate the extent to

which each individual node contributes to liquidity stress

amplification. Our contagion model shows that,in extreme

theoretical scenarios where liquidity buffers are small, a

handful of institutions may experiencesignificant spillover

effects due to the directionality of their portfolios.

The paper is organized as follows: Section 3 describes

the data and some descriptive statistics. Section 4 presents

the structural properties of the multiplex network and

compares the results coming from different centrality mea-

sures. Section 5 presents an analysis of the propagation of

VM shocks across the multiplex network, and Section 6

concludes the paper.

2|LITERATURE REVIEW

Multiplex networks (Bianconi, 2018) are a novel mathe-

matical framework that can capture the multilevel orga-

nization of a large variety of economic and financial

interactions. The strand of literature studying economic

and financial multiplex networks is relatively recent, but

fast‐growing. It often reveals important built‐in correla-

tions that can be used to extract information otherwise

unobtainable when aggregating across the layers. This

can have serious implications for the financial system.

For instance, Poledna et al. (2015) analyse the Mexican

banking system across four layers (derivatives, securities,

foreign exchange, and deposits and loans) and find that

focusing on single layers can lead to a drastic underesti-

mation of the total systemic risk by up to 90%.

Multiplex networks can be used to represent different

types of interaction between nodes (i.e., institutions). For

example, Berndsen et al. (2016) characterize the Colom-

bian financial system across three layers –the payment

system, the sovereign securities settlement system, and

the spot market foreign exchange –by analysing them both

independently and in aggregate. Meanwhile, Bookstaber

and Kenett (2016) describe the interplay among three

layers representing short‐term funding, assets, and collat-

eral flows. By focusing on the case of Bear Stearns during

the financial crisis, their analysis illustrates how risk can

spread from one layer to another. The Mexican and Italian

interbank markets are studied in Molina‐Borboa et al.

(2015) and Bargigli et al. (2015), respectively. Both build

multiplex networks whose layers correspond to different

types of contracts, and both find that properties such as

the persistence of lending relationships between the same

counterparties vary widely across different layers.

A common problem in the multiplex literature is how to

aggregate the information contained in the different layers.

Usually, this is archived by performing a weighted average

over the layers by assigning them a specific level of influ-

ence. Gould et al. (2018) build a network of countries inter-

connected across layers such as trade, migration, and

transportation systems. They calculate multiplex centrality

measures based on the works by Iacovacci et al. (2016) and

Rahmede et al. (2017) by assigning weights (or influences)

to each layer via a maximum likelihood procedure that tar-

gets economic growth. Because we do not have an obvious

quantity to target (e.g., optimal size of the interest rate

derivative market), we explore the full range of possible

weights that can be assigned to each layer and calculate

centrality measures by making the average of the centrality

measures for each weight combination (see Appendix B).

Korniyenko et al. (2018) consider a multiplex network in

which layers correspond to different kinds of investments

between countries. In their paper, they use a centrality