The impact of debt covenants on earnings announcement returns

ABSTRACT

We study the impact of debt covenants on earnings announcement returns, using event-study methodology, by creating 10 covenant groups and a covenant index. We find that during bad news, whether it stems from a bad earnings surprise based on analyst forecasts or a negative average market reaction, both the index and most covenant groups display a significantly negative impact on announcement returns. For good news, particularly when the market reaction is positive on average, we also observe a significantly positive effect from some covenant groups, but not the index. The strongest impact on returns comes mostly from debt-related covenants such as leverage tests, cross-default clause and debt issue restrictions. Finally, we also show that regardless of firm characteristics, risk measures or exposure to information asymmetry, most covenants have a negative impact on announcement returns for all firms when the news is bad. For the full sample and the good news subsample, however, we find that small, risky firms with high information asymmetry issues are rewarded by the market and vice versa.

KEYWORDS:

• Debt covenant
• earnings announcement
• equity valuation
• corporate governance

JEL CLASSIFICATION:

• G10
• G30
• G32

Acknowledgements

This work was supported by the BAGEP Award of the Science Academy.

Conflicts of interest

No potential conflict of interest was reported by the author(s).

Notes

¹ To conserve space, we present the results of this analysis in detail in section A.3 of the online appendix and discuss the main findings only here.

² See, for example, Chava and Roberts (2008), Beneish and Press (1993), Roberts and Sufi (2009), Falato and Liang (2016), and Nini, Smith, and Sufi (2009, 2012).

³ See, for example, Sweeney (1994), DeFond and Jiambalvo (1994), Dichev and Skinner (2002), Beatty, Ramesh, and Weber (2002), Franz, Hassabelnaby, and Lobo (2014) and Graham, Harvey, and Rajgopal (2005).

⁴ The individual covenants included in each category are listed and discussed in detail in section A.1 and Table A1 of the online appendix.

⁵ Our results are qualitatively robust to using alternative announcement windows such as (0,0), (0,1) and (−1,2).

⁶ Our results are robust to calculating market beta using daily returns from the past six months or simply subtracting the cumulative market returns from cumulative stock returns during the announcement window without a beta adjustment.

⁷ We also use the mean analyst forecast rather than the median and/or use the stock price at the end of the previous fiscal year as a standardizing variable to calculate SUE. Our results are robust to these choices.

⁸ In unreported results, we also include several other investor attention-related control variables such as Abnormal Volume, which is the trading volume on the day of the announcement minus the average daily trading volume during the last 3 months, and Turnover, which is the average ratio of trading volume to shares outstanding for the last 5 trading days including the announcement day. Finally, to capture institutional attention, we also use dummy variables for Tuesday through Friday since DellaVigna and Pollet (2009) suggest that the likelihood of institutional attention shocks decrease monotonically throughout the week. Our results are robust to the inclusion of all these control variables.

⁹ The split for analyst coverage is not even since the number of analysts following a firm is a discrete variable.

¹⁰ The correlation patterns and conditional probabilities related to the covenant categories are presented in Table A2 of the online appendix.

¹¹ Brandt et al. (2008) study the drift in returns of portfolios formed based on the stock price reaction around earnings announcements based on both SUE and EAR and they show that the strategy based on EAR produces a much higher annual return than the traditional SUE approach. They also show that the EAR and SUE strategies appear to be independent of each other which suggests that it is reasonable to use these two measures as separate information signals.

¹² It is important to emphasize that while two signals in the same direction may provide a cleaner information environment, it also leads to a loss in the number of observations hence a potential reduction of power for statistical tests.

¹³ Stock issuance restriction does not appear to follow this pattern and shows only a weakly significant negative impact for the reinforced negative information sample.

¹⁴ Starks and Yoon (1995) show that dividend increases create wealth effects consistent with cash flow signalling and are indeed followed by increases in capital expenditure for the subsequent three years hence the wealth effects surrounding the announcement is consistent with cash flow signalling. A positive average return to earnings announcement may similarly signal available cash flow as well as subsequent investment opportunities with possibly extra financing needs from debt issuance.

¹⁵ In section A.3 of the online appendix, we turn our attention to firm characteristics, including proxies of information asymmetry and risk measures, and investigate how the existence of both the covenant index and individual covenant groups affect the EAR both in the full sample as well as under good or bad earnings surprises. The results are presented in Table A3 of the online appendix.