Debt covenants and analysts’ information environment

ABSTRACT

This study investigates whether private information from financial covenants in syndicated loans affects analysts’ information environment. Specifically, we examine the relationship between the number of financial covenants and common information in analysts’ forecasts, defined as the common-to-total information ratio. We demonstrate that the common information in analysts’ forecasts increases significantly after loan initiation when loans impose more financial covenants. We also find that the effect is more pronounced for balance sheet-based covenants relative to income statement-based covenants. Our evidence suggests that analysts use financial covenant information in loan contracts to shape their forecast behavior.

KEYWORDS:

• Analysts’ forecasts
• financial covenants
• syndicated loans

Acknowledgments

The authors thank Suresh Radhakrishnan (the editor) and the anonymous referees for their helpful comments. All remaining errors are our own. We also acknowledge the helpful comments from participants at the 2017 Accounting Theory and Practice Conference & 2017 Asian Accounting Associations conference (Co-organized by TAA, JAA and KAA).

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. Chen and Martin (2011), for example, find that bank-affiliated analysts can improve their forecast accuracy by accessing superior information from the lending relationship.

2. The common-to-total information ratio measures the degree to which individual analysts’ forecasts incorporate private information (Barron et al. 1998). If an analyst collects more (less) private information, the common-to-total information ratio decreases (increases). For detailed measurements, see section 3.2.2.

3. Issuing corporate bonds and borrowing from banks are the two most important ways firms receive debt financing. According to Spiceland, Yang, and Zhang (2016), public debt represented only 17% of the outstanding debt (Houston and James 1996) and private debt accounts for 80% of corporate debt for their sample of large Compustat firms (Dichev and Skinner 2002).

4. In practice, covenants are divided into three broad categories. Besides financial covenants, the other two categories are affirmative covenants and negative covenants. The former requires the borrower to take certain actions, such as timely submission of financial information to the lender. The latter averts the borrower from taking certain actions, such as making excessive capital expenditures.

5. See the literature review of Armstrong, Guay, and Weber (2010).

6. Several studies, thus, discuss the effects of information asymmetry and accounting quality on the structure of debt contracts, such as interest rates, concentration (measured by the Herfindahl index), the proportion retained by the lead arranger, and the number of participant banks. These findings suggest that when borrowers’ financial reporting quality is higher, there are lower interest rates, lower proportions retained by the lead arranger, higher concentration, and a greater number of participant banks (Sufi 2007; Ball, Bushman, and Vasvari 2008; Bharath, Sunder, and Sunder 2008; Kim and Song 2011; among others).

7. Also see Nikolaev (2010) in Appendix A for examples regarding the use of accounting figure-based terms and conditions in debt contracts.

8. The reason is that the higher the uncertainty of future earnings, the greater is the volatility of future earnings, and this increases the difficulty of predicting future earnings. Because larger forecast errors will impair analysts’ reputation, they may act to lower their forecast failure (Barron et al. 2002a; Barron, Byard, and Kim 2002b).

9. This can also be supported by the contents of analysts’ research reports. For example, Previts et al. (1994) indicate that analysts care about firms’ major projects, including modernization, acquisition, expansion, divestiture, and restructuring plans. These projects are evaluated, and their estimated effects are used in forecasting future performance. Because major projects are often backed up by sufficient funds, firms may make loans.

10. Specifically, Lennox, Francis, and Wang (2012) indicate that the results using the Heckman procedures are highly fragile and sensitive to minor changes in model specification. In a related vein, Larcker and Rusticus (2010) argue that it is difficult to find appropriate instruments for the first-stage regression that are not correlated with the second-stage error term, and the Heckman approach with weak and/or partially endogenous instruments can produce more biased estimates than OLS methods.

11. Regarding the window, we generally follow Barron, Byard, and Yu (2008). However, the event date in Barron, Byard, and Yu (2008) is the earnings announcement date; hence, they use 45 days and 30 days as their pre-window and post-window, respectively. Because our event is the loan announcement, which is not as routine as earnings announcements, we use longer windows.

12. We notice that Liu and Natarajan (2012) indicating the observed forecast dispersion is affected by analysts’ strategic behavior and, on average, understates the dispersion of unmanaged forecasts. However, as per table 5 in their paper, when using their modified measure of dispersion to estimate proxies for consensus and public and private information precision in Barron et al. (1998), the conclusion of the consensus measure (that is, our common information in the forecast measure) is the same, although the magnitude is different. Therefore, we believe that even if the values differ, our main conclusion will remain unchanged if we decompose the dispersion.

13. Lennox, Francis, and Wang (2012) point out the importance of testing for multicollinearity when using the Heckman procedure. We find that the variance inflation factor (VIF) for IMR is less than 10 for the full sample, indicating that multicollinearity is not an issue. In columns (1) to (4) of Table 3, the VIFs for the IMR are 7.14, 7.15, 7.08, and 7.18, respectively.

14. Prior research finds mixed results in terms of the impact of RFD on information quality and quantity. For example, some research finds an increase in forecast dispersion in the post-RFD period (Bailey et al. 2003; Irani and Karamanou 2003; Mohanram and Sunder 2006), but Francis, Nanda, and Wang (2006) and Heflin, Subramanyam, and Zhang (2003) find no change in forecast dispersion.

15. We also follow the literature to use a two-year event window because it is unclear as to how long it takes for the implementation of RFD to have its full effect on the firm (Kirk and Vincent 2014). That is, we designate the reconstitution period of 1998 to 1999 as the pre-RFD period and 2001 to 2012 as the post-RFD period. Our results remain similar with these revised classifications.

16. Please see Rule 100(b)(2) in RFD, which sets out four exclusions from coverage.

17. Jorion, Liu, and Shi (2005), for example, find that the information content of credit ratings increases after RFD because credit rating agencies are one of the exclusions.

Additional information

Funding

The first author thanks the Ministry of Science and Technology of the Republic of China (Taiwan) for financial support of this research under contract numbers NSC 101-2410-H-155−024.