Relationship Trading in Over‐the‐Counter Markets

• Author: NORMAN SCHÜRHOFF,TERRENCE HENDERSHOTT,DMITRY LIVDAN,DAN LI
• Published date: 01 April 2020
• DOI: http://doi.org/10.1111/jofi.12864
• Date: 01 April 2020

THE JOURNAL OF FINANCE •VOL. LXXV, NO. 2 •APRIL 2020

Relationship Trading in Over-the-Counter

Markets

TERRENCE HENDERSHOTT, DAN LI, DMITRY LIVDAN,

and NORMAN SCH ¨

URHOFF∗

ABSTRACT

We examine the network of trading relationships between insurers and dealers in the

over-the-counter (OTC) corporate bond market. Regulatory data show that one-third

of insurers use a single dealer, whereas other insurers have large dealer networks.

Execution prices are nonmonotone in network size, initially declining with more deal-

ers but increasing once networks exceed 20 dealers. A model of decentralized trade in

which insurers trade off the benefits of repeat business and faster execution quanti-

tatively fits the distribution of insurers’ network size and explains the price–network

size relationship. Counterfactual analysis shows that regulations to unbundle trade

and nontrade services can decrease welfare.

INSURANCE COMPANIES ARE VITAL FOR risk sharing: In exchange for premium

payments, they compensate for loss, damage, injury, treatment, or hardship.

To provide this coverage, insurance companies invest in a variety of financial

assets. However, corporate bonds dominate their investment portfolios,

accounting for almost 70% of their investments with a total value close to

∗Terrence Hendershott is with the Walter A. Haas School of Business, University of California.

Dan Li is with the Board of Governors of the Federal Reserve System. Dmitry Livdan is with

the Walter A. Haas School of Business, University of California. Norman Sch¨

urhoff is with the

Faculty of Business and Economics and Swiss Finance Institute, University of Lausanne. He is

also a Research Fellow of the CEPR. Hendershott provides expert witness services to a variety of

clients. He gratefully acknowledges support from the Norwegian Finance Initiative. Li and Livdan

have no conflict of interest to disclose. Sch¨

urhoff gratefully acknowledges research support from

the Swiss Finance Institute and the Swiss National Science Foundation under Sinergia project

CRSII1_154445/1 “The Empirics of Financial Stability” and has no conflict of interest to disclose.

We thank Philip Bond; the Editor; an associate editor; two anonymous referees; Darrell Duffie;

Jerˆ

ome Dugast; Burton Hollifield; Haoxiang Zhu; and seminar audiences at California Polytechnic

University, Carnegie Mellon University, FED Board, Higher School of Economics Moscow, IDC

Herzliya, Imperial College, Johns Hopkins University, Michigan State University, National Bank

of Belgium, NY FED, Rice University,Texas A&M University,University of British Columbia, Uni-

versity of Houston, University of Illinois-Chicago, University of Oklahoma, University of Piraeus,

University of Wisconsin at Madison, University of Toronto,University of Victoria, WU Vienna, Ein-

audi Institute for Economics and Finance, 2015 Toulouse Trading and Post-Trading Conference,

2016 AFA, 2016 NBER AP Meeting, 2016 SFI Research Day, 12th Annual Central Bank of France

Workshop on the Microstructure of Financial Markets, 2016 LAEF Conference on OTC Markets,

2017 13th Csef-Igier Symposium on Economics and Institutions, and 2017 OTC Markets and their

Reform Conference in Rigi for valuable comments and feedback.

DOI: 10.1111/jofi.12864

C2019 the American Finance Association

683

684 The Journal of Finance R

$4 trillion.1To facilitate prompt compensation to policy holders, insurers need

to be able to liquidate their holdings quickly without incurring large trans-

action costs (Koijen and Yogo (2015), Chodorow-Reich, Ghent, and Haddad

(2016)). Yet, corporate bonds trade on decentralized over-the-counter (OTC)

markets, which are less liquid than centralized exchanges due to search fric-

tions arising from fragmentation and limited transparency (Duffie, Garleanu,

and Pedersen (2005,2007), Weill (2007), Vayanos and Wang (2007)). Insurers

therefore have to search for best execution across more than 400 active broker-

dealers. In this paper, we investigate whether insurers and other market

participants search randomly in the OTC corporate bond market, or whether

they build long-term relations with dealers to mitigate search frictions.

Regulatory data provide information on the transactions between more than

4,300 insurers and their dealers over the period 2001–2014. Using these data,

we begin by empirically examining insurers’ choice of trading network and the

relation between the network choice and transaction prices. Figure 1provides

examples of two different insurer-dealer trading networks. Panel A shows

buys and sells for an insurer that trades exclusively with a single dealer,

whereas Panel B shows buys and sells for an insurer that trades with multiple

dealers over time. We find that insurers tend to form small but persistent

dealer networks. At one extreme, approximately one-third of insurers trade

with a single dealer annually.At the other extreme, a small fraction of insurers

trades with up to 40 dealers each year. The overall degree distribution follows

a power law with exponential tail starting at about 10 dealers. When we

estimate trading costs as a function of network size N, we find that costs are

nonmonotone in N—costs decline with Nfor small networks and then increase

once Nexceeds 20 dealers.

Our results provide insights into which models of trade in OTC markets

better describe the empirical evidence on client-dealer networks and trading

costs. In random search models, clients repeatedly search for best execution

without forming a finite network of dealers (Duffie, Garleanu, and Pedersen

(2005,2007), Lagos and Rocheteau (2007,2009), Gavazza (2016)). The fact

that insurers form finite dealer networks suggests that adding dealers must

be costly for insurers. Traditional models of strategic search (e.g., Stigler

1961) assume that each additional dealer imposes a fixed cost on insurers, in

which case insurers add dealers to improve prices up to the point at which the

marginal benefit of doing so equals the fixed cost. However, this leads prices to

improve monotonically in network size, which is inconsistent with our finding

that trading costs are nonmonotone in network size.

To rationalize the empirical evidence, we build a model of decentralized

trade in which insurers—or more generally, clients—establish relationships

with multiple dealers and trade off the benefits of repeat business and faster

execution. We model the relationship between clients and dealers as having

1Insurers are major suppliers of capital to corporations. Schultz (2001) and Campbell and

Taks ler ( 2003) estimate that insurers hold between 30% and 40% of corporate bonds and account

for about 12% of trading volume.

Relationship Trading in OTC Markets 685

Figure 1. Two types of dealer-client trading networks. This figure shows the buy (squares)

and sell (circles) trades of two insurance companies with different dealers. We sort dealers on the

vertical axis by the first time they trade with the corresponding insurance company. (Color figure

can be viewed at wileyonlinelibrary.com)

two components. The first component captures the extent of repeat trading

between a client and a dealer in her network, which both the client and

the dealer take into account when negotiating the terms of a transaction.

The second component captures all client-dealer business unrelated to bond

trading. The latter component includes transactions in other securities, the

ability to purchase newly issued securities, as well as other soft dollar and

nonmonetary transfers such as investment research.

In our model, a single console bond trades on an interdealer market that

clients can access only through dealers. Dealers have search intensity λand

upon trading with a client transact at the competitive interdealer bid and ask

prices. Clients initially start without a bond but stochastically receive trading

shocks with intensity ηthat lead them to simultaneously contact Ndealers to

buy. Trading intensity η, and hence N, vary across clients. A client’s effective

search intensity, =Nλ, is increasing in the number of dealers. The first