Corporate Fraud and Corporate Bond Costs: Evidence from China

ABSTRACT

This study investigates the relationship between corporate fraud and four typical components of costs associated with corporate bonds. Based on data from a booming corporate bond market in China, we confirm that fraudulent issuers have higher corporate bond costs. Specifically, they are more likely to push upward price revisions, pay higher issue fees and coupon spreads, and encounter larger underpricing after issuance. Moreover, we demonstrate that severe corporate fraud is also significantly related to the costs of corporate bonds. Furthermore, we find that investors pay more attention to fraud in accounting information and disclosure. These results remain robust to a strand of endogeneity and through the robustness tests. In additional research, we find that bonds issued by fraudulent firms tend to receive lower ratings and show inferior performance after issuance. We also demonstrate that the effects of corporate fraud on bond costs erode as time passes, although the mitigation speed is slow. Finally, we find that hiring reputable financial intermediaries can partially mitigate the negative effects of corporate fraud.

KEY WORDS:

• corporate fraud
• cost of corporate bonds
• credit ratings
• financial intermediaries
• regulatory enforcement

Acknowledgments

All authors thank the constructional suggestions from the editor Ali Kutan and three expert reviewers.

Notes

^1. Filing coupons are calculated as the mid-value of the inquiry interval.

^2. Because some corporate bonds do not have trading records on the first listing day, we use the ending price of the first trading date to measure the underpricing level. However, we only keep those bonds with trading records within 7 days after formal listing. This procedure can exclude bonds that have high liquidity risk. We sincerely thank the referee for this suggestion.

^3. We also use the SSE Enterprise Bond Index to measure the underpricing level of corporate bonds. The correlation between the SSE Enterprise Bond Index and the CSI Enterprise Bond Index is 0.998. The untabulated regression based on the underpricing calculated by the SSE Enterprise Bond Index also shows that there is a significantly positive relationship between corporate fraud and underpricing.

^4. Notably, the CSRC requests that all Chinese corporate bonds be issued at their par value. Hence, the coupon rate can accurately measure the real financing cost of the bond issuer.

^5. SOEs are defined as borrowers directly or indirectly owned or controlled by state asset management bureaus or other SOEs controlled by the central government or local governments (Chen, Shi, and Xu 2013).

^6. The CSMAR and Wind databases were developed according to the international database standards to meet the requirements of academic research, and their databases are used in several recent studies. See, for example, Chen, Shi, and Xu (2013) and Chen and Zhu (2013).

^7. As suggested by Ge and Kim (2014), we use lagged accounting variables, which has two advantages. First, the offering yield is more affected by past accounting information than by current accounting information, which ensures that accounting information is already available to bondholders at the time of bond issuance. Second, regression of current bond yields on lagged accounting information alleviates a potential endogeneity concern.

^8. According to Du et al. (2017), we also employ the condition indices to diagnose the multicollinearity. Nontabulated results show that the largest intercept-adjusted condition index is far less than 10, suggesting no serious multicollinearity in our empirical models (Belsley 1991).

^9. We should admit that, as indicated by Conyon and He (2016), a better solution would be instrumental variables. However, the difficulty here is that it is problematic to find a legitimate theoretical instrument that is correlated with the propensity to commit fraud (the relevant criterion) and is also uncorrelated to those cost variables. In consequence, any chosen instrument set might turn out to be theoretically somewhat arbitrary.

^10. The results based on the extended PSM subsample are similar. These results are available from the authors upon request.

^11. In untabulated analysis, we also calculate accrual quality by using methods of Dechow and Dichev (2002) and Ball and Shivakumar (2006). However, we do not find any evidence to contradict these findings. These results are available upon request.

^12. The NERI index is constructed by using principal components analysis based on scores on five aspects. Here, we simply adopt the principal index values. We measure the regional market development level using the lagged NERI index at the end of the fiscal year before bond issuance. However, since Fan, Wang, and Zhu (2011) provide data for the NERI index across various regions in China from 2001 to 2009 only, we use the regional NERI index measured in 2009 for bonds issued after 2010. Additionally, following Amit et al. (2010), we also use a dummy variable as the alternative measurement for the regional market development level, which equals one if the NERI value of the market of the province in which the firm is headquartered is greater than the median and zero otherwise. We repeat our regressions and find that the results are consistent with the main results. For brevity, the results are not tabulated.

^13. In China’s corporate bond market, most rating agencies only provide unsolicited rating services. This means that even for listed firms, they can only receive formal bond and firm ratings when they prepare to issue corporate bonds. The data structure of credit ratings cannot support an event study following Bonini and Boraschi (2010). Therefore, we adopt a regression analysis and acknowledge this limitation.

^14. We should admit that using the ordinal variable to measure credit rating is with deficiencies because the gaps between rating levels probably are different. However, we cannot use a strand of dummies here because the credit rating is the dependent variable in the mediation path. We acknowledge this limitation here. We sincerely thank the referee’s comments on this issue.

^15. We also conduct path analyses using Firmrating as the mediation variable, which is measured similarly to Creditrating to proxy the firm rating. The results are still consistent. For brevity, these results are untabulated.

^16. Ideally, we should use the default events to denote the performance of corporate bonds. However, there are only limited default events in China’s corporate bond market today. Hence, we alternatively use rating changes to demonstrate bond performance.

^17. We also adopt another strategy to examine this inference. We include the interaction terms of fraud variables and time range between detected fraud events and bond issuance. The coefficients of interaction terms are significant and negative, which also support our prediction. These results are not tabulated, but they are available upon request.

^18. These results are not tabulated, but they are available from the authors upon request.

Additional information

Funding

All authors acknowledge the funding support from the Ph.D. Programs Foundation of Ministry of Education of China (Grant number 20130161110045).