Informed Trading in Corporate Bonds Prior to Earnings Announcements

• Published date: 01 August 2016
• Author: Xing Zhou,Jason Wei
• DOI: http://doi.org/10.1111/fima.12123
• Date: 01 August 2016

Informed Trading in Corporate Bonds

Prior to Earnings Announcements

Jason Wei and Xing Zhou∗

This paper examines the information content of corporate bond trading prior to earnings

announcements using data from both NAIC and TRACE. We find that the direction of pre-

announcement bond trading is closely related to earnings surprises. This link is most evident

prior to negative news and in high-yield bonds. Further, abnormal bond trading during the pre-

announcement period can help predict both earnings surprises and post-announcement bond

returns. Such predictive ability of bond trading largely originates from institutional-sized trades

and is concentrated in the issuer’s most actively traded bond. Finally, even after accounting for

transactions costs, informed bond tradingcan generate significant net profits, especially prior to

the release of bad news.

Common stocks and corporate bonds are traded securities representing claims on the same

underlying assets of the issuing firm. Investors in either market with superior information on

the firm’s cash flows can potentially benefit from initiating trades with the uninformed. Such

information advantage is especially valuable prior to corporate events, because information in the

imminent news can allow informed investors to establish favorable positions before significant

price changes.

Several studies have used the announced earnings to gauge the arrival of new information and

examined potential informed trading in the equity market. Although Lee (1992) finds no evidence

of informed trading in the buy-sell patterns of equity traders during the day before earnings

announcement, most studies provide evidence consistent with informed trading in relatively

longer pre-announcement windows. For example, based on a sample of large earnings surprises,

Affleck-Graves, Jennings, and Mendenhall (1994) find that the direction of large trades during

the 30 days prior to the announcement largely anticipates the actual announcements. Seppi

(1992) shows that price changes of block trades prior to earnings announcements reflect private

information about unexpected earnings. Christophe, Ferri, and Angel (2004) find that abnormal

pre-announcement short selling is linked to post-announcement stock returns. Finally, Kaniel

et al. (2012) show that stock trading by individual investors can predict earnings sur prises.

In contrast, little attention has been paid to the potential informed trading in the corporate

bond market. This is not surprising as bonds are traditionally suited for buy-and-hold strategies.

Moreover,liquidity of the bond market is much lower than the equity markets, and bonds are more

expensive to trade than stocks for general investors (see, e.g., Chacko, 2006; Edwards, Harris,

and Piwowar, 2007). Further, bond investors do not enjoy higher leverage, such as that offered in

We thank Edith Hotchkiss, Simi Kedia, Tavy Ronen, Daniel Weaver, and participants at Baruch College and Rutgers

University seminars and the 2012 China International Conference in Financefor helpful comments. We especially thank

the Editor,Marc Lipson, and an anonymous refereefor extensive comments and suggestions. Weacknowledge the financial

support from the Social Sciences and Humanities Research Council of Canada.

∗Jason Weiis at Department of Management, University of Toronto Scarborough & Joseph L. Rotman School of Man-

agement, University of Toronto, Toronto,Canada. Xing Zhou is with Federal Reserve Board, Washington,DC.

Financial Management •Fall 2016 •pages 641 – 674

642 Financial Management rFall 2016

the options markets. Therefore, corporate bonds are usually not considered as a typical vehicle

for informed trading.

However, unlike the equity market that has active participation of individual investors, the

corporate bond market is mainly dominated by large institutions. Recent studies (e.g., Barber

et al., 2009; Ivashina and Sun, 2011; Boulatov, Hendershott, and Livdan, 2013) have shown that

trading by these sophisticated investors in the equity market tends to be information based. In the

bond market, even if a typical institutional investor (e.g., an insurance company) largely pursues

a buy-and-hold strategy, it can still benefit by strategically timing the trades (necessitated by,

e.g., portfolio rebalancing) around pubic news releases such as earnings announcements. This

is especially the case when the information in the pending announcements can cause significant

changes to the return of their fixed income portfolios.

Furthermore, because the issuer’scash-flow prospect directly affects the credit risk in corporate

bonds and earnings are an accounting indication of cash flows, information on earnings (once

publicly released) can exert nontrivial impact on bond values. In fact, by studying a sample of

large quarterly earnings surprises, Datta and Dhillion (1993) find that earnings announcements

have information content for both the stock and the bond markets. In sum, plausible arguments

can be made both for and against informed trading in corporate bonds. In particular, whether

pre-announcement bond trading is informative about the to-be-announced earnings remains an

open question.

In this paper, we focus on the information content of trading activities in the corporate bond

market prior to the actual earnings announcements. Our main findings are the following. First,

even though corporate bonds do not trade often, when they do trade, these trades seem to be

informative. Wefind elevated level of bond trading in the 10 days leading up to the announcement.

In our TRACE sample, the median number of trades/trade volume in a firm’s bonds during the

10-day pre-announcement period is 23/$8.96 million, which is much higher than 12/$5.11 million

during the non-announcement period on the same 10-day basis. More importantly, the direction

of pre-announcement trading is significantly related to the pending earnings surprise, especially

when the surprise is negativeand large in magnitude. This link remains signif icant after weaccount

for the potential influence of recent earnings forecasts, abnormal price movements in the issuer’s

equity,the f irm’s accounting information, and the general market and macroeconomic conditions.

Our results are also robust to firm and year f ixed effects, alternative pre-announcement window

sizes, alternative ways of standardizing the earnings surprises, and different bond datasets.

Second, pre-announcement informed trading is mainly observed in high-yield bonds, especially

before large negative earnings surprises. According to Merton (1974), corporate debt can be

valued as a combination of a long position in a risk-free debt and a short position in a put option

on the issuer’s assets. Therefore, the moneyness of the option will determine the relevance of

earnings information for bond traders. For a higher-rated bond, the put option is most likely deep

out-of-the-money (because the default risk is low), and earnings news is of little consequence.

However, earnings information becomes highly valuable for lower-rated bonds, especially when

a large negative surprise is pending. Informed traders would most likely trade risky bonds prior

to negative news, a prediction confirmed by our results.

Third, wef ind that pre-announcement bond trading can help predict both earnings surprises and

post-announcement bond returns. The prediction is evident in both univariate and multivariate

analyses. Excess abnormal selling in bonds, especially in high-yield bonds, predicts negative

returns following the release of bad news. Such predictive power remains after we incorporate

the information from trading and price movements in the issuer’s equity. In addition, we find that

the trade imbalance from institutional-sized trades exhibits significant power in predicting returns,

while the retail-sized trades exhibit little predictive power. Further, the ability of large trades in

Wei & Zhou rCorporate Bonds Prior to Earnings Announcements 643

predicting returns varies across bond issues by the same firm, and it is mostly concentrated in the

issuer’s “top bond”—the bond with the highest trading volume by institutional investors.

Finally, pre-announcement bond trading, especially prior to the release of bad news, can be

quite profitable. This is so even after accounting for the relatively high transactions costs in

corporate bond trading. We find that selling prior to bad news can generate over 100 basis

points in returns within a 10-day holding period. Such profits are net of transaction costs and are

after adjusting for contemporaneous returns in rating- and maturity-matched bond indexes. No

statistically significant profits are found from buying prior to good news.

Our paper’s main contributions can be summarized as follows. First, to the best of our knowl-

edge, this paper is the first to present a thorough empirical study of informed trading in corporate

bonds prior to earnings announcements. The potential gains from informed trading in the bond

market seem to outweigh transactions costs. Although informed traders might not buy bonds

prior to good news, they are quite alert on bad news, and tend to sell bonds prior to the public

announcements of this news. Our paper therefore complements the studies on informed trading

in the equity market prior to earnings news, and provides a broader picture of the information

revelation process concerning corporate earnings. It also contributes to the growing literature on

the general efficiency and quality of the bond market. For example, Das, Kalimipalli, and Nayak

(2014) find that, contrary to common perceptions, the introduction of the CDS market actually

made the bond market less efficient.

Second, our research sheds light on the multi-venue trading on firm specific information and

on the relative information efficiency of stocks and bonds. Earlier studies rely on the vector

autoregression (VAR) modes to determine the lead-lag relationship between stocks and bonds in

incorporating firm specif ic information, and found conflicting results.1The special institutional

features of the corporate bond market have led Ronen and Zhou (2013) to question the ability

of the VAR approach in capturing the dynamics of information revelation and argue in favor

of event studies for cross-market comparisons. By examining the information content of bond

trading prior to earnings announcements, our study complements Ronen and Zhou (2013) and

suggests that VAR and pair-wise comparisons between the time-series of bond and stock prices

might not be effective in fully revealing the relative informational efficiency of the stock and

corporate bond markets. The time-series analyses ignore the asymmetrical nature of informed

bond trading prior to good and bad news.

Finally, our study contributes to the literature on the relevance of earnings information for the

credit markets. Empirical studies have documented significant bond market reactions to earnings

announcements (Datta and Dhillion, 1993; Hotchkiss and Ronen, 2002; Ronen and Zhou, 2013).

By focusing on the information content of trading activities before earnings announcements, our

study suggests that the earnings impact on bond prices might have been evenlarger than previously

documented, because the trading by informed traders has incorporated some information into the

prices prior to the public announcements. What is more, we show that trading activities in the

corporate bond market can actually predict earnings surprises. This finding can potentially open

a new line of research on earnings predictions.

It is useful to note the relationship between our study and the literature on bond yield

spreads. Bond liquidity (Chen, Lesmond, and Wei, 2007; Kalimipalli, Nayak, and Perez, 2013),

firm’s internal liquidity (Chen, Liao, and Tsai, 2011), earnings forecast dispersion (G¨

untay and

Hackbarth, 2010), and information uncertainty (Lu, Chen, and Liao, 2010) have all been shown

1For example,using weekly quote data, Kwan (1996) finds that stocks lead bonds in revealing fir m-specific information,

a conclusion also reached by Downing, Underwood, and Xing (2009). However, Hotchkiss and Ronen (2002) find no

causal relation between stock and bond returns in a sample of liquid high-yield bonds.