Benchmarks in Search Markets

• Author: HAOXIANG ZHU,PIOTR DWORCZAK,DARRELL DUFFIE
• Published date: 01 October 2017
• DOI: http://doi.org/10.1111/jofi.12525
• Date: 01 October 2017

THE JOURNAL OF FINANCE •VOL. LXXII, NO. 5 •OCTOBER 2017

Benchmarks in Search Markets

DARRELL DUFFIE, PIOTR DWORCZAK, and HAOXIANG ZHU∗

ABSTRACT

We characterize the role of benchmarks in price transparency of over-the-counter

markets. A benchmark can raise social surplus by increasing the volume of benefi-

cial trade, facilitating more efficient matching between dealers and customers, and

reducing search costs. Although the market transparency promoted by benchmarks

reduces dealers’ profit margins, dealers may nonetheless introduce a benchmark to

encourage greater market participation by investors. Low-cost dealers may also in-

troduce a benchmark to increase their market share relative to high-cost dealers. We

construct a revelation mechanism that maximizes welfare subject to search frictions,

and show conditions under which it coincides with announcing the benchmark.

AN ENORMOUS QUANTITY OF OVER-THE-COUNTER (OTC) trades are negotiated by

counterparties who rely on the observation of benchmark prices. In this paper

we explain how benchmarks affect pricing and trading behavior by reducing

market opacity, we characterize the welfare impact of benchmarks, and we

show how the incentives of regulators and dealers to support benchmarks

depend on market structure.

Trillions of dollars in loans are negotiated at a spread over LIBOR or EURI-

BOR, benchmark interbank borrowing rates. LIBOR is the London Interbank

Offered Rate. EURIBOR is the Euro Interbank Offered Rate. For U.S. dollar

LIBOR alone, the Market Participants Group (MPG) on Reference Rate Re-

form (2014) (chaired by one of the authors of this paper) reports that over

∗Darrell Duffie is with Stanford University Graduate School of Business and NBER. Piotr

Dworczak is with Stanford University Graduate School of Business. Haoxiang Zhu is with MIT

Sloan School of Management and NBER. Weare grateful for helpful discussions with and comments

from Bruno Biais (Editor), two anonymous referees, Ana Babus, David Bowman, Gregory Connor,

Willie Fuchs, Will Gornall, Brett Green, Terry Hendershott, Gustavo Manso, Konstantin Milbradt,

Paul Milgrom, Jose Moraga-Gonzalez, Marzena Rostek, Ali Shourideh, Andy Skrzypacz, Chester

Spatt, Jeremy Stein, Gabor Virag, and Xavier Vives,as well as seminar and conference participants

at Stanford University, Princeton University, U.C. Berkeley, Harvard University, MIT, Finance

Theory Group, SFS Cavalcade, European Central Bank, Paul Woolley Centre Annual Conference,

Barcelona GSE Summer Forum, Western Finance Association annual meeting, Bonn Workshopon

Information Aggregation, Society for Economic Dynamics meeting, NBER Summer Institute Asset

Pricing meeting, Econometric Society World Congress, Philadelphia Fed Search and Matching

Conference, the 11th Central Bank Workshop on Microstructure of Financial Markets, Conference

on Monetary Policy Implementation and Transmission in the Post-Crisis Period, and American

Finance Association annual meeting. We also thank members of the Market Participants Group

on Reference Rate Reform (MPG) for useful discussions and comments. Duffie was appointed chair

of the Market Participants Group on Reference Rate Reform by the Financial Stability Board. The

authors have read the Journal of Finance’s disclosure policy and have no conflict of interest to

disclose.

DOI: 10.1111/jofi.12525

1983

1984 The Journal of Finance R

3 trillion dollars in syndicated loans and over 1 trillion dollars in variable-rate

bonds are negotiated relative to LIBOR. The report of the Market Participants

Group lists many other fixed-income products that are negotiated at a spread

over the “interbank offered rates” known as LIBOR, EURIBOR, and TIBOR,

across five major currencies. As of the end of 2013, the Bank for Interna-

tional Settlements (2014) reports a total notional outstanding of interest rate

derivatives of 583 trillion U.S. dollars, the vast majority of which reference

LIBOR or EURIBOR. These swap contracts and many other derivatives refer-

ence benchmarks but are not themselves benchmark products. Other extremely

popular benchmarks for overnight interest rates include SONIA, the Sterling

OverNight Index Average, and EONIA, the Euro OverNight Index Average.

The WM/Reuters daily fixings are the dominant benchmarks in the foreign

exchange market, which covers over $5 trillion per day in transactions.1There

are also popular benchmarks for a range of commodities including silver, gold,

oil, and natural gas, among others.2Benchmarks are additionally used to pro-

vide price transparency for manufactured products such as pharmaceuticals

and automobiles.3

Among other roles, benchmarks mitigate search frictions by lowering the

informational asymmetry between dealers and their “buy-side” customers. We

consider a market for an asset in which dealers offer price quotes to customers

who are relatively uninformed about the typical cost to dealers of providing the

asset. We provide conditions under which adding a benchmark to an opaque

OTC market can improve efficiency by encouraging entry by customers, im-

proving matching efficiency, and reducing total search costs.

Recent major scandals over the manipulation of benchmarks for interest

rates, foreign currencies, commodities, and other assets have made the robust-

ness of benchmarks a major concern of international investigators and policy

makers. This paper offers a theoretical foundation for the public policy support

of transparent financial benchmarks. In Section IV we discuss the manipula-

tion of benchmarks in more detail.

Our model works roughly as follows. In an OTC market with a finite number

of dealers and a continuum of investors that we call “traders,” the cost to a

dealer of providing the asset to a trader is the sum of a dealer-specific (idiosyn-

cratic) component and a component that is common to all dealers. (In practice

the clients of financial intermediaries may be buying or selling the asset. We

consider the case in which traders wish to buy. The opposite case is effectively

the same, up to sign changes.) The existence of a benchmark is taken to mean

1See Foreign Exchange Benchmark Group (2014), which reports that 160 currencies are covered

by the WM/Reuters benchmarks. These benchmarks are fixed at least daily and by currency pair

within the 21 major “trade” currencies.

2The London Bullion Market Association provides benchmarks for gold and silver. Platts pro-

vides benchmarks for oil, refined fuels, and iron ore (IODEX). Another major oil price benchmark

is ICE Brent. ICIS Heren provides a widely used price benchmark for natural gas.

3For a discussion of the Average Wholesale Price (AWP) drug price benchmarks, see Gencarelli

(2005). The Kelly Blue Book publishes the “Fair Purchase Price” of automobiles based on the

average transaction price by model and location.

Benchmarks in Search Markets 1985

that the common cost component is publicly announced. Each trader privately

observes whether her search cost is high or low. Traders are searching for a

good price, and dealers offer them price quotes that depend endogenously on

the presence of a benchmark. Each dealer posts an offer price, which is avail-

able for execution by any trader, anonymously. Traders, who have a commonly

known value for acquiring the asset, contact the dealers sequentially, expend-

ing a costly search effort or costly delay with each successive dealer contacted.

At each point in time, the trader, given all of the information available to her

at that time (including past price offers and, if published, the benchmark),

decides whether to buy, continue searching, or exit the market. All market

participants maximize their conditional expected net payoffs, at all times, in a

perfect Bayesian equilibrium.

Under natural parameter assumptions, which vary with the specific result,

we show that publishing the benchmark is socially efficient because of three

types of effects.

First, publication of the benchmark encourages efficient entry by traders,

thus increasing the realized gains from trade. The benchmark improves the

information available to traders about the likely price terms they will face.

This assists traders in deciding whether to participate in the market, based on

whether there is a sufficiently large conditional expected gain from trade. The

increased transparency of prices created by the benchmark induces dealers

to compete more aggressively in their quotes. In this sense, publication of

the benchmark mitigates the hold-up problem caused by dealers’ incentives to

quote less attractive prices once the search costs of traders have been sunk.

Second, benchmarks improve matching efficiency, which leads to a higher

market share for low-cost dealers. When the benchmark is not observed by

traders, high-cost dealers exploit the ignorance of traders about the cost of

providing the asset and may conduct sales despite the presence of more efficient

competitors. The benchmark allows traders to decompose a price offer into a

common-cost component and a dealer-specific component for cost and profit

margin. As a result, if search costs are sufficiently small, customers trade with

the most efficient dealers.

Finally,benchmarks reduce wasteful search by (i) alerting traders that gains

from trade are too small to justify entry, and (ii) helping traders infer whether

they should stop searching because they have likely encountered a low-cost

dealer.

We also characterize cases in which the introduction of a benchmark low-

ers welfare. This can happen when the market is already relatively efficient

without the benchmark.4

We embed the price transparency problem—add a benchmark or not—into

a broader design framework by characterizing a socially optimal revelation

mechanism. Here, we take the case in which dealers have the same costs.

We show that whenever the gain from trade between a dealer and a trader

4This finding is consistent with the insight of Asriyan, Fuchs, and Green (2015) (in a very

different model) that welfare can be nonmonotone in the degree of transparency.