Over‐the‐Counter Market Frictions and Yield Spread Changes

• Published date: 01 December 2019
• DOI: http://doi.org/10.1111/jofi.12827
• Date: 01 December 2019
• Author: NILS FRIEWALD,FLORIAN NAGLER

THE JOURNAL OF FINANCE •VOL. LXXIV, NO. 6 •DECEMBER 2019

Over-the-Counter Market Frictions and Yield

Spread Changes

NILS FRIEWALD and FLORIAN NAGLER∗

ABSTRACT

We empirically study whether systematic over-the-counter (OTC) market frictions

drive the large unexplained common factor in yield spread changes. Using transac-

tion data on U.S. corporate bonds, we find that marketwide inventory, search, and

bargaining frictions explain 23.4% of the variation in the common component. System-

atic OTC frictions thus substantially improve the explanatory power of yield spread

changes and account for one-third of their total explained variation. Search and bar-

gaining frictions combined explain more in the common dynamics of yield spread

changes than inventory frictions. Our findings support the implications of leading

theories of intermediation frictions in OTC markets.

ACCORDING TO FRICTIONLESS NO-ARBITRAGE THEORY, changes in corporate yield

spreads occur because of innovations in firm-specific and macroeconomic funda-

mentals. This paradigm has been challenged by empirical studies showing that

yield spread changes are difficult to explain. Using conventional factors, Collin-

Dufresne, Goldstein, and Martin (2001, CDGM henceforth) show that a large

set of firm-specific and macroeconomic variables perform poorly in explaining

the variation in yield spread changes over time. A substantial proportion of the

unexplained variation is due to a single common factor.

∗Nils Friewald is at Norwegian School of Economics and CEPR. Florian Nagler is at Bocconi

University. We thank INQUIRE Europe for financial support, as well as Ali´

e Diagne and Ola

Persson of the Financial Industry Regulatory Authority (FINRA) for providing us access to a

proprietary data set comprising transactions in the U.S. corporate bond market. We thank Stefan

Nagel (Editor) and two anonymous referees for many insightful comments and suggestions. We

are grateful for comments from Hendrik Bessembinder, Paolo Colla, Jean-Edouard Colliard, Jesse

Davis, Narayan Naik, Jens-Dick Nielsen, Joost Driessen, Carlo Favero, Arvind Krishnamurthy,

Spencer Martin, Sebastian M ¨

uller,Nicola Gennaioli, Walter Pohl, Thomas Kjær Poulsen, Francesco

Saita, Francisco Santos, Andriy Shkilko, Marti G. Subrahmanyam, Christian Wagner, Michael

Weber,participants at the 2014 VGSF Conference, the 2015 SFS Finance Cavalcade, the Seventh

Erasmus Liquidity Conference, the 22nd Annual Meeting of the German Finance Association, the

25th Anniversary Seminar of INQUIRE Europe, the 2016 Annual Meeting of the American Finance

Association, the 20th Annual Conference of the Swiss Society for Financial Market Research,

the 2017 European Summer Symposium in Financial Markets, as well as seminar participants

at Copenhagen Business School, Norwegian Business School, Norwegian School of Economics,

Luxembourg School of Finance, and Vienna University of Economics and Business for helpful

comments and suggestions. We have read the Journal of Finance’s disclosure policy and have no

conflicts of interest to disclose. Any remaining errors are our own.

DOI: 10.1111/jofi.12827

3217

3218 The Journal of Finance R

U.S. corporate bonds trade in an over-the-counter (OTC) market in which sev-

eral dealers manage bond inventories to provide liquidity to customers. Trans-

actions are nonanonymous and occur on a bilateral basis and thus the terms of

the trade are determined by search and bargaining frictions. Consequently,the

theoretical literature starting with Duffie, Gˆ

arleanu, and Pedersen (2005,2007,

DGP henceforth) rationalizes deviations of prices from fundamentals through

OTC market frictions. In this paper, we empirically investigate the ability of

time-varying OTC frictions to explain the remaining common component of

yield spread changes. We find that systematic inventory, search, and bargain-

ing frictions explain 23.4% of the variation in the common component and

account for one-third of the total explained variation in yield spread changes.

To establish our findings, we employ detailed transaction data on the prices

and volumes of U.S. corporate bonds. Specifically, we use a version of the Trade

Reporting and Compliance Engine (TRACE) database that contains dealer in-

formation that allows us to assign every transaction to a particular dealer. The

final data set captures all trades executed by more than 2,600 dealers over

the 2003 to 2013 sample period. Our transaction data reveal similar patterns

as the quote- and price-based data of CDGM. In particular, when we imple-

ment the CDGM baseline regression model on our sample of 974 bonds, we

find that the explanatory power of monthly yield spread changes is low, with

a mean adjusted R2of 21.7%. Using principal component analysis (PCA), we

further find that the residuals are highly cross-correlated and exhibit a large

common component. That is, the first principal component captures 48.4% of

the unexplained variation.

After having established the CDGM benchmark result, we investigate

whether the common component of yield spread changes is related to sys-

tematic OTC frictions. Several studies show that proxies for transaction costs

relate positively and systematically to yield spread changes (e.g., Longstaff,

Mithal, and Neis (2005), Chen, Lesmond, and Wei (2007), Bao, Pan, and Wang

(2011)). However, while transaction costs are symptomatic of intermediation

frictions, previous literature does not provide insights into the type of friction

that affects yield spread changes. Intermediation frictions are hard to mea-

sure, which renders their empirical investigation just as challenging. We fill

this gap by exploiting the granularity of our data set to construct proxies for

the intensity of systematic inventory, search, and bargaining frictions in the

corporate bond market.

First, we focus on the role of systematic inventory frictions. Theories based

on Stoll (1978)andHoandStoll(1981,1983) relate asset prices to dealer

inventories. These theories predict that an increase in the level of aggregate

dealer inventory reduces prices (increases yield spreads) and vice versa. We use

aggregate order flow to proxy for changes in marketwide inventory. More ad-

vanced theories of liquidity provision by financially constrained intermediaries

further imply that increases in dealers’ time-varying risk aversion and in their

funding costs of holding inventory,respectively, reduce asset prices (Gromb and

Vayanos (2002), Brunnermeier and Pedersen (2009), Nagel (2012)). To measure

dealers’ risk aversion, we use the fact that dealers can avoid inventory risk

Over-the-Counter Market Frictions and Yield Spread Changes 3219

by prearranging trades between sellers and buyers and conjecture that more

prearranged trades imply more risk-averse dealers. We use the TED spread

to proxy for dealers’ funding costs. In yield spread regressions, we find that

all measures are significant, exhibit the predicted sign, and reduce the com-

mon unexplained variation across bonds. Specifically, we find that marketwide

inventory frictions jointly explain 13.9% of the variation of the common com-

ponent.

Second, we examine the effect of systematic search frictions on yield spread

changes. The broad implication of the random search framework of DGP and

Lagos and Rocheteau (2009) is that asset valuations increase when search

frictions relax and thus counterparties are easier to find. However, Di Maggio,

Kermani, and Song (2016) provide empirical evidence indicating that dealers

do not randomly search but rather form trading networks to mitigate search

frictions, consistent with models of network formation (e.g., Neklyudov (2014),

Chang and Zhang (2016), Wang (2016)). Our first search proxy therefore is a

measure of the overall connectivity between dealers, which we define as the

graph-level eigencentrality of the interdealer network. As expected, we find

that yield spreads narrow when dealers are more closely connected, implying

that bonds are easier to locate and search frictions are lower.

Next, recent theory on intermediation chains in OTC markets (e.g.,

Hugonnier, Lester, and Weill (2016), Shen, Wei,and Yan (2016), Neklyudov and

Sambalaibat (2017)) suggests that the intensity of search frictions is reflected

in the properties of intermediation chains. Accordingly, we identify intermedi-

ation chains by tracing bonds through the interdealer network after they have

been sold by customers and before they disappear into clients’ portfolios. We

allow for split intermediation chains, which generalize the chains introduced

by Hollifield, Neklyudov, and Spatt (2017) and Li and Sch¨

urhoff (2019)inthat

there are multiple sales to dealers along the chain. We first focus on the chain

length, that is, the number of dealers involved in the chain. We find that longer

chains are associated with smaller yield spreads. This result is consistent with

the models of, for example, Shen, Wei, and Yan (2016) and Neklyudov and

Sambalaibat (2017), where lower search costs lead to an endogenously larger

intermediary sector with more competitive allocations and thus longer chains.

The theory of intermediation chains also models search frictions as in DGP

using the meeting rate between sellers and buyers. In any intermediation

chain, an initial volume is disseminated to a number of final customers. We

therefore use the chains as a laboratory to posit conclusions about meeting

rates, that is, we investigate the required number of sales to customers to com-

plete the chain. We allow for heterogeneity and examine chains with large and

small initial volumes separately.We find that more sales in large-volume chains

result in smaller yield spreads. This result suggests that when dealers split

large initial volumes, search frictions are low and meeting rates are high, as

dealers can rely on a large client base. In contrast, for small-volume chains, we

find that more sales lead to a widening of yield spreads. This finding indicates

that when dealers disseminate a relatively small volume to many customers,