Corva › Blog › Covenant headroom
Debt and covenants
An investor who divides net debt by trailing EBITDA and calls the gap covenant headroom has computed the wrong ratio, on purpose, from the lender's point of view. Here is how to find the real one, and rebuild it.
10 min read· Corva Research
The short version
Covenant headroom is the room a borrower has left before a leverage or coverage covenant breaks. Screeners approximate it by dividing net debt by trailing EBITDA and comparing the result to a threshold pulled from a bond prospectus. That approximation is usable for a first pass and wrong for a second one.
It is wrong because the lender's definition of EBITDA is not the number sitting in a data provider's cash flow statement. Credit agreements build their own version, with an addback stack that commonly runs 20% to 40% above the GAAP-derived figure. Restructuring charges, stock compensation, transaction costs, non-cash items, and often a forward-looking line for cost savings that have not happened yet. The direction of the error is always the same. Covenant EBITDA is bigger, so real headroom is smaller than a naive calculation suggests, until the addbacks run out.
The second thing almost nobody does: read the sensitivity a company already gave you. Issuers under real covenant pressure frequently state, in their own MD&A, how far EBITDA could fall before they breach. That sentence is a free answer key, and this article shows two companies that published one.
Covenant headroom is the distance between a borrower's actual financial ratio and the limit set in its credit agreement or bond indenture, expressed as either a turn of leverage or a percentage decline in EBITDA the borrower could absorb before tripping the covenant. It is a narrower question than whether a company has too much debt overall; a business can carry a conservative debt load and still sit close to a covenant if the covenant itself is tight.
Three ratio types dominate leveraged credit agreements. A maximum total leverage ratio caps total debt against trailing EBITDA. A maximum secured leverage ratio does the same for secured debt only, and is usually tighter. A minimum interest coverage ratio requires EBITDA to exceed interest expense by some multiple. All three move the same way when EBITDA falls: the leverage ratios rise toward their ceiling, and the coverage ratio falls toward its floor.
Headroom shrinks faster than it looks, because leverage is a ratio, not a difference. A company running 7.0x against a 7.75x covenant is not "0.75 turns safe" in any linear sense. A modest EBITDA decline moves the denominator enough to close that gap, as the worked example below shows.
The covenant itself is rarely hard to find. The definition behind it is. Both live in predictable places, in this order. If you have not sat down with a 10-K before, our guide to reading a 10-K in 30 minutes covers where Item 7 sits and how the rest of the document is organised.
That last item is the one almost everyone skips. The 10-K tells you the covenant exists and whether the company passed it. It almost never tells you what counts as EBITDA for the test. For that, you need the exhibit.
Open a real one. CorePoint Lodging's credit agreement, filed as Exhibit 10.13 to its FY2019 Form 10-K, defines Consolidated EBITDA starting from Consolidated Net Income and then adding back, among other items:
"the amount of any restructuring charges or reserves, equity-based or non-cash compensation charges or expenses including any such charges or expenses arising from grants of stock appreciation or similar rights, stock options, restricted stock or other rights, retention charges… start-up or initial costs for any project or new production line, division or new line of business or other business optimization expenses or reserves… plus any other non-cash charges, including any write-offs or write-downs… plus (i) the amount of ‘run-rate’ cost savings, operating expense reductions and synergies… that are projected by the Borrower in good faith to result within 24 months… from actions that have been taken or with respect to which substantial steps have been taken or are expected to be taken… calculated on a pro forma basis as though such cost savings… had been realized on the first day of such period."
Read that last clause twice. The lender lets the borrower add back savings that have not been realised, based on actions the borrower merely expects to take, projected two years forward, as if they had already happened on day one of the measurement period. No percentage cap appears in this clause. That is not an edge case. It is standard drafting.
Academic evidence confirms how common this is. A 2022 Federal Reserve Bank of St. Louis working paper parsed EBITDA definitions from 4,112 real credit agreements and found non-GAAP addbacks in all but 344 of 3,939 loan packages, roughly 91%. About 43% carried three or more distinct addback categories. The five recurring categories, in the paper's own classification:
The same paper cites an S&P Global study of merger and buyout transactions, finding that issuers' projected adjusted EBITDA at deal inception exceeded the EBITDA they actually realised in the following two years by about 30% on average. The addback stack does not just widen a definition. It moves a forecast into a covenant test as though it were a fact.
Here is the mechanism, using round numbers rather than any one company's figures, so the arithmetic is easy to follow. Start from reported EBITDA and walk up the addback stack a typical credit agreement allows.
| Line item | $M |
|---|---|
| Reported EBITDA (operating income plus D&A) | 100.0 |
| + Restructuring charges and reserves | 4.5 |
| + Equity-based and non-cash compensation | 3.0 |
| + Other non-cash charges and write-downs | 2.5 |
| + Transaction and integration costs | 3.0 |
| + Projected run-rate cost savings and synergies | 12.0 |
| = Covenant-defined Consolidated EBITDA | 125.0 |
Addback total: $25.0 million on $100.0 million of reported EBITDA, a 25% uplift. That sits inside the 20% to 40% range documented across real credit agreements. Source: addback categories drawn from CorePoint Lodging's Credit Agreement, Exhibit 10.13 to its FY2019 Form 10-K, filed 13 March 2020, and from Faria-e-Castro, Gopalan, Pal, Sánchez and Yerramilli, "EBITDA Add-backs in Debt Contracting: A Step Too Far?", Federal Reserve Bank of St. Louis Working Paper 2022-029, 2022.
An investor computing a leverage ratio off the $100.0 million line gets a materially worse number than the lender's own covenant test. That gap is not an error. It is the entire point of negotiating the definition, and it is the reason a naive calculation always reads tighter than the covenant actually is, right up until the addbacks stop being available.
Boyd Gaming's Form 10-K for fiscal year 2010, filed with the SEC on 15 March 2011, shows what this looks like with real, disclosed numbers rather than an illustration. At 31 December 2010, the company reported three ratios under its Amended Credit Facility against three separate limits.
| Covenant | Actual | Limit | Decline to breach |
|---|---|---|---|
| Total Leverage Ratio (max) | 7.07x | 7.75x | 8.7% |
| Secured Leverage Ratio (max) | 4.21x | 4.50x | 6.4% |
| Interest Coverage Ratio (min) | 2.84x | 2.00x | 29.5% |
Source: Boyd Gaming Corporation, Form 10-K for the year ended 31 December 2010, filed 15 March 2011, Item 7, Liquidity and Capital Resources.
The company stated its own sensitivity plainly:
"At December 31, 2010, assuming our current level of Consolidated Funded Indebtedness remains constant, we estimate that an 8.7% or greater decline in our twelve-month trailing Consolidated EBITDA, as compared to December 31, 2010, would cause us to exceed our maximum permitted consolidated Total Leverage Ratio covenant for that period."
Check the arithmetic yourself, because a table that does not add up is caught by the exact reader you most want to convince. Leverage is debt over EBITDA, so if EBITDA falls by a fraction x, the ratio becomes 7.07 ÷ (1−x). Set that equal to the 7.75x ceiling and solve: 1−x = 7.07 ÷ 7.75 = 0.9123, so x = 8.77%. That matches the 8.7% Boyd disclosed. The same method on the secured ratio, 4.21 ÷ 4.50, returns 6.4%. On interest coverage, where a falling EBITDA pushes the ratio down instead of up, 2.84 × (1−x) ≥ 2.00 solves to x ≤ 29.6%, matching the disclosed 29.5%. All three of Boyd's own numbers reconcile from two disclosed ratios and nothing else.
Notice what those figures are not. They are declines in Consolidated EBITDA as defined in the Amended Credit Facility, not in the segment-level "Adjusted EBITDA" of $459.8 million that the same 10-K reports elsewhere for property performance. The two numbers are built for different purposes and are not required to match. An investor who pulls the $459.8 million figure and treats an 8.7% decline in it as the covenant test is applying the disclosed sensitivity to the wrong base.
Entravision Communications' Form 10-K for fiscal year 2008, filed 16 March 2009, shows the same discipline applied by a different company in a different industry, a year defined by falling advertising revenue rather than falling gaming volumes.
At 31 December 2008, Entravision's net debt ratio stood at 5.2x, up from 4.9x at the end of both 2007 and 2006. The company remained in compliance, and it told investors exactly how much room was left:
"if our trailing-twelve-month consolidated adjusted EBITDA were to decrease in excess of between approximately 29% to 32% from consolidated adjusted EBITDA for the year ended December 31, 2008, the Company believes that it would breach the maximum allowed leverage ratio covenant."
Entravision went further than Boyd. In the same section it disclosed the revenue decline that would produce that EBITDA outcome: a 15% to 25% quarterly revenue drop, repeated across 2009, would put the covenant at risk. That is a rare thing to find: a company translating a leverage covenant all the way down into a top-line number a reader can actually forecast.
The procedure, end to end. First, locate the covenant and its limit in Item 7. Second, find the credit agreement exhibit and read the EBITDA definition, not just its name. Third, rebuild covenant EBITDA from the reported figure using the addback categories the agreement actually lists, not a generic assumption. Fourth, compute the leverage or coverage ratio on that rebuilt figure. Fifth, and this is the step almost nobody adds: compare your own number against any sensitivity the company disclosed itself.
Three outcomes are possible on that last step, and only one of them is comfortable.
None of this replaces reading the actual definition. It is a way to know, within a few minutes, whether your estimate is in the right neighbourhood before you trust it.
Here is the honest limit, and it undercuts the whole exercise above if you let a headroom number sit unrefreshed. A covenant headroom figure computed from an annual filing is stale the moment it is published.
A headroom calculation is a snapshot of a moving target, taken from the slowest camera available. Treat it as a floor on your understanding of the risk, not a forecast of what will happen next quarter.
A promise the borrower makes to its lenders to keep a financial ratio, usually debt to EBITDA or EBITDA to interest expense, inside an agreed range. Breaching it lets lenders demand repayment or renegotiate terms, even if every payment was made on time.
In the credit agreement itself, filed as an exhibit to a 10-K or 8-K, searchable on SEC EDGAR by company name and exhibit type EX-10. The 10-K body rarely reproduces the full definition, only the covenant's existence and a compliance statement.
Because the addback stack is negotiated at the same table as the interest rate and the covenant level. A borrower accepts a wider EBITDA definition in exchange for a covenant it can live inside; the lender prices the resulting risk into the spread.
No. Companies disclose it voluntarily, typically when leverage is elevated enough that investors would ask anyway. Its absence does not mean headroom is fine. It usually means nobody has forced the company to say.
Not by itself. Wide addback stacks are the market norm, present in roughly 91% of the loan packages studied in the Federal Reserve research cited above. The useful question is how large they are relative to reported EBITDA, and whether run-rate synergies are doing most of the work.
No. Screener EBITDA is standardised for comparability across companies. Covenant EBITDA is bespoke, built for one purpose: passing one company's specific test. The two numbers answer different questions and are not substitutes for each other.
Corva reads the filing, locates the debt covenants management discusses, and flags where a disclosed sensitivity exists so you are not rebuilding it from scratch. Figures are computed from the filed statements and cross-checked against SEC EDGAR. Where a number cannot be found, it says so rather than inventing one.
Research a company free →Live financials on any listed company are free. No card.
Corva is a research tool, not a broker or investment adviser, and nothing here is a recommendation to buy, sell or hold any security. Boyd Gaming Corporation and Entravision Communications Corporation are used here as documented examples of disclosed covenant sensitivity and for no other reason; the illustrative bridge table is hypothetical and does not describe either company's actual figures. Boyd Gaming figures are from its Form 10-K for the year ended 31 December 2010, filed 15 March 2011. Entravision figures are from its Form 10-K for the year ended 31 December 2008, filed 16 March 2009. Verify anything you intend to act on against the primary filing and the current credit agreement, not this article. See terms and disclaimer.