Measurement Error when Estimating Covenant Violations
59 Pages Posted: 18 Nov 2021
Date Written: September 27, 2021
Prior research often estimates covenant slack and violations using financial data from Compustat and covenant thresholds from Dealscan, which may induce measurement error since covenants are commonly based on non-GAAP numbers and Dealscan often fails to capture threshold changes associated with contractual amendments. Using a hand-collected sample of actual covenant thresholds and realizations from SEC filings, we find that estimated covenant slack measures grossly overstate the frequency of covenant violations. Although Type I errors (overestimation) are common, Type II errors (underestimation) are rare, suggesting the measurement error is not random. We show Type I errors are responsible for previous findings that suggest lenders fail to “enforce” approximately 90% of covenant violations (Bird et al. 2020). We identify differences between true covenant realizations and estimated realizations from Compustat data as the primary source of measurement error and propose measurement adjustments to reduce measurement error when estimating covenant slack.
Keywords: Debt covenants, Covenant Slack, Covenant Violation
JEL Classification: G30, M40, M41
Suggested Citation: Suggested Citation