An Experimental Study of Bond Market Pricing

• Author: MATTHIAS WEBER,ARTHUR SCHRAM,JOHN DUFFY
• Published date: 01 August 2018
• DOI: http://doi.org/10.1111/jofi.12695
• Date: 01 August 2018

THE JOURNAL OF FINANCE •VOL. LXXIII, NO. 4 •AUGUST 2018

An Experimental Study of Bond Market Pricing

MATTHIAS WEBER, JOHN DUFFY, and ARTHUR SCHRAM∗

ABSTRACT

An important feature of bond markets is the relationship between the initial public

offering (IPO) price and the probability that the issuer defaults. On the one hand,

the default probability affects the IPO price; on the other hand, the IPO price affects

the default probability. It is a priori unclear whether agents can competitively price

such assets. Our paper is the first to explore this question. Todo so, we use laboratory

experiments. Wedevelop two flexible bond market models that are easily implemented

in the laboratory. We find that subjects learn to price the bonds well after only a few

repetitions.

NEW ASSETS ARE OFTEN BROUGHT to market using an initial public offering (IPO).

While a variety of studies analyze different IPO mechanisms, little is known

about the impact of IPO prices on the subsequent financial health of the com-

pany issuing the new assets. The IPO price can, for example, affect the perfor-

mance of an equity-issuing company through its effect on the funds available

to the company for additional investments. Similarly,with new debt issues, the

IPO price determines the financing costs of the bond issuer and in turn affects

the probability that the bond issuer is able (or willing) to meet new or preex-

isting debt obligations. Such feedback effects from IPO prices to the real side

of the economy and to the value of the assets themselves are often overlooked

in the literature on asset pricing and are fully absent from the experimen-

tal part of this literature. Our paper seeks to fill this void by conducting the

first experimental study of assets with endogenous, price-dependent default

probabilities.

∗Matthias Weber is with the research center CEFER at the Bank of Lithuania and with the

Faculty of Economics and Business Administration at VilniusUniversity (from August 2018, Weber

will be with the Swiss Institute of Banking and Finance at the University of St. Gallen). John

Duffy is with the Department of Economics at the University of California, Irvine. Arthur Schram

is with the Robert Schumann Center for Advanced Studies at the European University Institute

(Florence) and with the Amsterdam School of Economics at the University of Amsterdam. Wethank

Junze Sun for research assistance in running some of the experiments reported here. Thanks for

comments and suggestions go to Stefan Nagel (the Editor); a coeditor; an Associate Editor; two

referees; Tibor Neugebauer; Fr´

ed´

eric Schneider; and participants in the Experimental Finance

Conference in Mannheim, the SAEe Meetings in Bilbao, the ESA World Meetings in San Diego,

the WEAI Annual Conference in San Diego, and seminars in Amsterdam and Vilnius. Financial

support by The Netherlands’ Organization for Scientific Research (NWO grant 406-11-022), the

Research Priority Area Behavioral Economics of the University of Amsterdam, and the UC Irvine

School of Social Sciences is gratefully acknowledged. The authors have no conflicts of interest to

disclose. Disclaimer: The views expressed are those of the authors and do not necessarily reflect

those of the Bank of Lithuania.

DOI: 10.1111/jofi.12695

1857

1858 The Journal of Finance R

We focus on the pricing of bonds with credit risk, for two reasons. First, the

feedback from IPO prices to the probability that the bond issuer defaults and

hence to the value of the bonds is particularly clear in this case. While there is

evidence that default risk accounts for much of the yield spread between high-

risk corporate bonds and risk-free government bonds of comparable maturities

(Longstaff, Mithal, and Neis (2005), Huang and Huang (2012)), the interde-

pendency between the initial issue price of such bonds and default risk has not

been clearly identified. Second, bond markets dwarf markets for other assets

such as equities in terms of the value of outstanding issues (Roxburgh, Lund,

and Piotrowski (2011)). Indeed, bond sales are often the cheapest means of

financing compared with alternatives such as equity issues or bank loans, and

bond issues are the primary means of financing for most large, mature firms

and governments. This preference for bond financing presumes, however, that

bond purchasers are able to correctly price new bond issues. If newly issued

bonds are mispriced relative to fundamentals, the financing costs to the bond

issuer may be higher, leading to an increase in the issuer’s default risk.

Rather than use field data to address the pricing of bonds, we employ an

experimental approach where subjects in a laboratory are asked to buy new

risky bond issues and earn money proportionate to the payoffs possible in the

bond market setting in which they operate. An important advantage of a labo-

ratory evaluation of bond market pricing is that we can control the information

available to bond purchasers regarding features of the bond issuance, and in

particular the probability that the bond issuer defaults, and hence we can eval-

uate the extent to which this information allows them to correctly determine

bond prices. Knowledge of such information and this high level of control more

generally are not available or feasible in the field. For this reason, a large lit-

erature uses experimental methods to evaluate the pricing of many types of

assets (see, e.g., Bossaerts (2009) for a survey).

Specifically, we develop and evaluate the empirical relevance of two models

of bond pricing that follow the approach of Merton (1974), in that IPO prices are

related to features of the bond issue (e.g., face value, coupon payment, maturity

date) and to the probability that the bond issuer defaults on its obligations to

bondholders. In the first experiment, the relationship between prices in the

IPO and default probabilities is made as salient as possible to participants.

The model used in that experiment reduces the mechanisms behind the feed-

back from IPO prices to default probabilities to a single mapping, with default

probabilities that are equal in all periods. The second experiment is based on

a structural model that fully spells out the relationship between the revenue

raised in the IPO and the ability of the bond issuer to meet its obligations. In

this case, the default probabilities are not equal in different periods, and thus

no mapping such as that in the model of the first experiment exists. As a conse-

quence, participants in the second experiment need to derive all implications

of the feedback from IPO prices to default probabilities themselves.

To preview our results, we find that subjects learn to price bonds subject to

endogenous default risk rather well after only a few repetitions (both during

the IPO and while trading in the secondary market that follows). As in other

An Experimental Study of Bond Market Pricing 1859

asset pricing experiments, we find that bubbles in bond prices are only ob-

served among inexperienced traders. This is a remarkable finding, given that

the bonds in our experiments are relatively long-lived, giving ample opportu-

nity for bubbles to occur (e.g., Noussair and Tucker (2013)), and that learning

the equilibrium IPO price is not an easy task—we had to compute the solution

numerically. The fundamental value of the bond in the secondary market de-

pends on the IPO price, so that learning in the secondary market is related to

learning during the IPO. The overall large degree of market efficiency that we

find occurs in a variety of environments. In particular, we show that it occurs in

environments where the relationship between IPO prices and default probabil-

ities is straightforward (Experiment I), independent of whether (equilibrium)

fundamental values are increasing or decreasing over trading periods, and it

similarly occurs in environments where the relationship between IPO prices

and default probabilities is more complex (Experiment II), independent of the

structure of noise in the underlying cash flows.

Notwithstanding a large literature on experimental asset pricing,1we are not

aware of any prior experimental study that examines the pricing of bond-type

assets with interdependencies between the IPO price and default risk. Some

experimental asset markets contain an explicit possibility of default (e.g., Ball

and Holt (1998), Crockett, Duffy, and Izhakian (2017)), and in other asset

market experiments a random cash flow can be interpreted as a type of partial

default (e.g., Plott and Sunder (1988)), but in those studies the pricing of the

asset does not affect the probability of default as in our framework. While our

models can easily be implemented in the laboratory, this does not mean that

the computation of equilibrium prices is easy (in fact, these computations are

rather complex).

Recent surveys of the experimental asset pricing literature by Bossaerts

(2009), Noussair and Tucker (2013), and Palan (2013) reveal just a few exper-

iments that explicitly consider assets referred to as “bonds”; more generally

assets are simply labeled as “assets.” In the studies that do refer to certain as-

sets as bonds, the bonds are included primarily to expand the portfolio of assets

available to traders. For instance, in a study of risk-versus ambiguity-aversion,

Bossaerts et al. (2010) allow agents to trade in riskless bonds that pay a known

fixed return in contrast to other assets that have either known probabilistic

(risky) returns or ambiguous returns. Fischbacher,Hens, and Zeisberger (2013)

allow agents to trade in a riskless, interest-bearing bond as well as a Smith,

Suchanek, and Williams (1988) type risky asset. The latter is known to exhibit

price bubbles and crashes and so the authors’ aim is to explore the effect of

interest rate (i.e., monetary) policy (via the riskless bond) on the extent of mis-

pricing of the risky asset. By contrast, our focus in this paper is on bond pricing

1Recent studies include Gneezy, Kapteyn, and Potters (2003), Dufwenberg, Lindqvist, and

Moore (2005), Haruvy and Noussair (2006), Bossaerts, Plott, and Zame (2007), Haruvy, Lahav,

and Noussair (2007), Bossaerts et al. (2010), Palan (2010), Cheung and Palan (2012), Kirchler,

Huber, and St¨

ockl (2012), Sutter, Huber, and Kirchler (2012), Cheung, Hedegaard, and Palan

(2014), F¨

ullbrunn, Rau, and Weitzel (2014), Bosch-Rosa, Meissner, and Bosch-Dom`

enech (2018),

and Crockett, Duffy, and Izhakian (2017).