Financial Statement Anomalies in the Bond Market

Abstract

We investigate the association between bond returns and 32 financial statement variables. Our findings show that 17 of the 32 financial statement measures we examined are significantly related to future bond returns. Evidence of inefficiency is more pronounced when institutional investors are less active and when there is more uncertainty about the creditworthiness of the issuer. We contribute to the literature by significantly expanding the number of anomalies analyzed and by providing practitioners with actionable guidance on which trading strategies may be profitable in the bond market.

Disclosure: The authors report no conflicts of interest.

Authors’ Note

The data used in this study are available from public sources identified in the article.

Editor’s Note

Submitted 16 April 2018

Accepted 15 January 2019 by Stephen J. Brown

Acknowledgments

The authors are grateful to Giuseppe Ballochi, CFA, Stephen Brown, and Larry Pohlman for helpful guidance and to Brian Akins, Jeremiah Green, John Hand, Bradley Lail, Alfred Wagenhofer, and workshop participants at Utah State University, Brigham Young University, the 2013 EAA annual meeting, the 2013 Western Regional AAA meeting, the 2013 AAA annual meeting, the 2016 BAFA annual meeting, the 2016 EUFIN meeting, and the 2017 GBATA conference. Pietro Perotti gratefully acknowledges the financial support of the Austrian Science Fund (P 24911-G11).

This article was externally reviewed using our double-blind peer-review process. When the article was accepted for publication, the authors thanked the reviewers in their acknowledgments. Giuseppe Ballochi, CFA, and Larry Pohlman were the reviewers for this article.

Notes

¹ Note that in terms of total dollars invested, the bond market is significantly larger than the stock market. SIFMA (Securities Industry and Financial Markets Association) reported that as of the fourth quarter of 2017, the total value of the US bond market was $40.8 trillion, $8.8 trillion of which corresponds to corporate bonds; in the same period, the value of the US equity market, as measured by the capitalization of stocks listed in the Wilshire 5000 Total Market Index, was $27.8 trillion.

² See Table 2 for a list of the anomalies in each group.

³ Note that three of the variables (change in employees, industry-adjusted change in employees, and size) are not obtained using financial statement data. We included change in employees and industry-adjusted change in employees because investments in human capital are an important part of a company’s investment strategy not captured by the other financial statement investment variables. We included size because it has been shown to be a robust predictor of returns and is correlated with financial statement measures that capture the scale of the company.

⁴ We provide the details behind the sample construction in the online supplemental material, available at www.tandfonline.com/doi/suppl/10.1080/0015198X.2019.1572377.

⁵ We also replicated the analysis using decile portfolios instead of quintile portfolios. We chose to use quintiles in the main analysis to ensure a larger number of observations in the hedge portfolio. The results using deciles are generally similar to those in the main analysis.

⁶ Table S3 in the online supplemental material (available at www.tandfonline.com/doi/suppl/10.1080/0015198X.2019.1572377) presents the mean values for each fundamental variable in the long and short portfolios.

⁷ Another approach for calculating characteristic-adjusted returns for each company is as the difference between the return for the company and the return of a portfolio matched on specific characteristics. Jostova et al. (2013) documented similar inferences for the two approaches, but the regression approach allows for the simultaneous adjustment of multiple factors.

⁸ Readers can contact the corresponding author, Pietro Perotti ([email protected]).

⁹ Using intraday data from Enhanced TRACE over 2004–2012, Schestag et al. (2016) estimated that the average relative spread is 1.2864%. We used this estimate in our calculations. We calculated turnover as the percentage of companies that are in a given portfolio in a month and were not in the portfolio in the prior month. The mean turnover for the long and short portfolios formed on each of the fundamental variables is available in Table S3 of the online supplemental material (www.tandfonline.com/doi/suppl/10.1080/0015198X.2019.1572377). Turnover in most of the portfolios from one month to the next was less than 10%.

¹⁰ Our proxy for institutional trading has a mean of 24.38% and a standard deviation of 0.30. The median, which is used to define the partitions, is 11.76%. Note that the proportion of large trades is lower than the mean reported by Even-Tov (2017)—36.06%—perhaps because he considered earnings announcement windows, where trading is likely to be substantially more active.

¹¹ The median of the absolute earnings surprise, which was used to define the partitions, is 0.0202.