Explaining repo specialness

• Published date: 01 April 2020
• Author: Ivan Sangiorgi,Frank S. Skinner,Miriam Marra,Alfonso Dufour
• DOI: http://doi.org/10.1002/ijfe.1746
• Date: 01 April 2020

RESEARCH ARTICLE

Explaining repo specialness

Alfonso Dufour

1

| Miriam Marra

1

| Ivan Sangiorgi

1

| Frank S. Skinner

2

1

ICMA Centre, Henley Business School,

University of Reading, Whiteknights

Campus, Reading, UK

2

Kingston Lane, Brunel University London,

Uxbridge, Middlesex, UK

Correspondence

Ivan Sangiorgi, Lecturer in Finance at the

ICMA Centre, Henley Business School,

University of Reading, Whiteknights

Campus, Reading RG6 6BA, UK.

Email: ivan.sangiorgi@icmacentre.ac.uk

Abstract

We study the dynamics of specialness for 1-day repo contracts on Italian govern-

ment bonds over a 10-year sample period. As predicted by Duffie's (1996) model,

our results show that collateral supply is a significant factor for specialness. How-

ever, we enrich that finding by also showing a clear impact from repo liquidity,

collateral riskiness, information uncertainty, and short-selling proxies, revealing the

importance of speculative bond demand for specialness. During crisis periods, bond

fire sales and European Central Bank interventions also have a large impact on repo

specialness. We identify recurrent patterns for specialness around bond auctions.

Specialness increases steadily from the auction announcement date until a few days

before the auction settlement date, which is consistent with overbidding behaviour

and a short selling of treasuries (via reverse repos) from primary dealers ahead of

auctions.

KEYWORDS

auctions, fire sales, high-frequency data, liquidity, repo specialness, short selling

JEL CLASSIFICATION

E43; E51; G01; G12; G15; G24

1|INTRODUCTION

The aim of this paper is to explain the variation in the degree

of specialness of repurchase agreements (repos) that use Ital-

ian government coupon bonds (BTPs—Buoni del Tesoro

Pluriennali) as collateral. Repos are financial instruments for

collateralized borrowing and they are essential for well-

functioning and efficient bond markets. Repos are used by

bond market participants to either finance long bond posi-

tions or initiate short bond positions. Repo markets are also

a preferred monetary transmission channel used by central

banks to conduct money market operations and regulate

financial market liquidity.

1

Repo markets are characterized

by huge transaction volumes, as repor ted in Table 1, which

presents data for the Mercato dei Titoli di Stato (MTS). The

MTS is one of the largest European electronic bond and

repo markets and the largest one for Italian government

bonds (Dunne, Moore, & Portes, 2006). During our sample

period from April 1, 2003 to December 6, 2013, the overall

value and the daily average value ofall repo transactions exe-

cuted on the MTS are about €179 trillion and €65 billion,

respectively.

In a general collateral (GC) repo, the borrower can spec-

ify the securities they want to use as collateral after the repo

trade is agreed. The collateral is chosen from a predefined

basket of treasuries. In contrast, in a special collateral repo,

the borrower specifies a unique security as collateral at the

outset of a repo transaction and pays a (generally) lower

“special”repo interest rate on the loan. At each point time,

the difference between the general and the special collateral

repo rates measures the “degree of specialness”of the bond

used as collateral in the special collateral repo. Positive

DOI: 10.1002/ijfe.1746

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

provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

© 2019 The Authors. International Journal of Finance & Economics published by John Wiley & Sons Ltd

Int J Fin Econ. 2019;1–25. wileyonlinelibrary.com/journal/ijfe 1

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any

medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.

© 2019 The Authors. International Journal of Finance & Economics published by John Wiley & Sons Ltd

172 Int J Fin Econ. 2020;25:172–196.wileyonlinelibrary.com/journal/ijfe

specialness is generally considered as a signal of greater

“market desirability”or the relatively scarce supply of the

specific instrument used as collateral in the repo contract. In

this paper, we provide novel empirical evidence on Italian

government bonds and develop a comprehensive model for

explaining repo specialness. We also observe that the degree

of specialness changes over time and across bonds, and often

exhibits some recurrent patterns that are correlated with the

bond auction cycles.

The first theoretical underpinning of our empirical inves-

tigation on Italian government bonds repo specialness is

provided by the model developed by Duffie (1996) for the

U.S. repo market. Duffie (1996) explains that specialness

can arise when collateral owners are inhibited from supply-

ing the collateral in repos because of legal/institutional

requirements or frictional/opportunity costs. Most of the

existing literature testing Duffie's model of repo specialness

focuses on U.S. government bond special repos. Jordan and

Jordan (1997) empirically test Duffie's model using data on

overnight repos that have U.S. treasuries as collateral. They

find that the lower supply of collateral for repos (measured

by higher auction tightness and lower ownership by dealers)

causes higher repo specialness. Several other studies in-

vestigate the U.S. Treasury repo market (see for instance

Fisher, 2002; Graveline & McBrady, 2011; Keane, 1995;

Krishnamurthy, 2002; Moulton, 2004; and Sundaresan,

1994). They explain the degree of specialness using long-

term treasury supply as the main explanatory variable, but

also control for auction cycles, liquidity demand, on- and

off-the-run status of the bonds, and some other risk factors.

More recently, D'Amico, Fan, and Kitsul (2018) focus on

the effects of Federal Reserve Board (FED) programs on

U.S. overnight special repo rates during the period

2009–2013 by looking at the amounts of treasuries pur-

chased and sold by the FED.

2

Few papers have studied repo specialness in the

European markets. Buraschi and Menini (2002) find that

long-term German sovereign special repos overestimate the

future relative scarcity value of bonds on special because of

the existence of time-varying liquidity risk premiums.

Corradin and Maddaloni (2017) analyse the Italian sovereign

overnight repo market over the short period October 2009 to

July 2012, which includes the European sovereign debt crisis

(ESDC). They show that the European Central Bank (ECB)

interventions, in particular, the outright purchases of govern-

ment bonds in the Securities Markets Programme (SMP)

framework, had an important effect on the repo markets by

affecting the supply of collateral.

We look at a much longer sample period, from April

2003 to December 2013, covering the tranquil period of

April 2003–August 2007; the 2007–2009 global financial

crisis (GFC); the 2010–2012 ESDC; and the post-crisis

period of February 2012–December 2013. We use a dataset

of intraday bond data and repo daily data from MTS. This

allows us to run a comprehensive test of Duffie's (1996) the-

oretical predictions for the Italian government bond repo

market, to examine whether a larger set of risk factors than

those indicated by Duffie's model have affected specialness

over this extended period and whether/how the effects of

some key variables on specialness have changed over time.

3

Furthermore, our empirical work analyses three different

repo-term contracts: overnight (ON1), tomorrow-next (TN),

and spot-next (SN) repos. We provide a detailed analysis of

the dynamics of specialness for these three different repo

terms over the Italian government bonds' auction cycles. This

comparative analysis has never been reported in either

U.S. or European repo studies and it reveals some interesting

differences. We find that repo contracts on Italian BTPs

exhibit persistent and significant patterns that are related to

the reopening of existing bond issues. With a reopening auc-

tion on day T, the amount of bonds outstanding increases

starting from the auction settlement date T+ 2 onwards. We

observe that, on average, specialness tends to increase

steadily for all repo terms from the announcement date of the

reopening until a few days before the actual auction settle-

ment date, and then it decreases. The effect of auctions on

specialness tends to decrease over consecutive reopening

auctions and it varies across bond maturities. The pattern of

repo specialness around auctions varies across the three repo

terms, being consistent with their contractual differences in

the timing of collateral exchange and with higher short sell-

ing activity by dealers via reverse-repos ahead of auctions.

Dealers hedge the risk they are about to acquire at auction by

short selling similar securities in the secondary market before

the auction. We define this behaviour as hedging the

TABLE 1 Nominal amount of transacted collateral in the MTS

repo market

General

Collateral

(Euro

Billions)

Special

Collateral

(Euro

Billions)

Total (Euro

Billions)

Daily

average

28.90 36.57 65.45

Standard

deviation

9.54 11.27 13.98

Minimum 0.75 0.06 0.06

Maximum 99.28 130.66 189.53

Trading days 2,735 2,735 2,735

Total 79,035.66 100,044.69 179,080.35

Note. The table reports summary statistics for the nominal volume of collateralin

MTS general and special repo transactions over the period April 1, 2003 to

December 6, 2013.

2DUFOUR ET AL.

DUFOUR ET AL.173