What did Houlihan Lokey's third-quarter loan data show about coverage?
The first decline in over a year, against rising earnings. Median borrower interest coverage fell to 1.68 times in September from 1.73 times in June, which the release calls "the first decline since mid-2025", and the share of borrowers with coverage below 1.0 times rose to 15.1 percent from 13.9 percent (Houlihan Lokey, Q3 2026 Private Credit DataBank, released 19 November 2026).
What makes that print worth an owner's attention is what it happened alongside. Median earnings did not fall. The release reports median revenue up 5.9 percent and median earnings up 8.1 percent year over year, with 70 percent of borrowers growing revenue and 66 percent growing earnings, and most borrowers "at or near peak performance": 85 percent sit at or within 10 percent of their highest observed revenue. Coverage still slipped, because the cost of the debt moved against a book that was otherwise improving.
The findings come from the firm's Q3 2026 Private Credit DataBank Market Trends and Insights report, which the release states is "available to clients of the firm". The DataBank is built from Houlihan Lokey's own private credit valuation work, carries observations back to 2017 and more than 300 standardized fields on each loan, and spans "the full direct lending market, from borrowers with less than 20 million dollars of EBITDA to those with more than 100 million" (same release). Everything below is from the release itself, not from the client report.
Which does more damage, a rate rise or an earnings fall?
The earnings fall, by roughly two to one on the measure that triggers covenants. Houlihan Lokey ran both tests on the same book and states the conclusion plainly: "Interest coverage sensitivities show coverage is more dependent on earnings than rates" (same release).
The two legs are not close. A 100 basis point rise in base rates "would lower median coverage to 1.55x and lift the sub-1.0x share to 18.0%". A 25 percent decline in earnings at current rates "would cut the median to 1.26x and double the sub-1.0x share to 30.0%" (same release). Measured against the September starting point, the rate shock moves the median by 13 hundredths of a turn and the earnings shock by 42. On the share of the book that stops covering its interest at all, the rate shock adds under three percentage points and the earnings shock adds almost fifteen.
In our judgement that asymmetry is the most useful thing in the release for an owner, and it runs against the way borrowers are usually encouraged to think about risk. The rate is the variable that gets watched, because it is printed daily and because it is the one a lender will offer to help manage. The earnings line is the variable that actually decides whether a coverage test is met, and it is the one nobody sells a product against.
It is worth being precise about what these two figures are. They are not forecasts, and the release does not present them as a view on rates or on the economy. They are the arithmetic of the existing book under two stated shocks, which is why they are comparable to each other: the same loans, the same starting coverage, one variable moved at a time.
| Case | Median interest coverage | Share of borrowers below 1.0 times |
|---|---|---|
| Reported, June 2026 | 1.73x | 13.9% |
| Reported, September 2026 | 1.68x | 15.1% |
| Base rates 100 basis points higher, earnings unchanged | 1.55x | 18.0% |
| Earnings 25 percent lower, rates unchanged | 1.26x | 30.0% |
Why does the earnings leg fall hardest on smaller borrowers?
Because the swing is wider there, in both directions. Lower middle market borrowers, which the release defines as those with between 10 and 20 million dollars of EBITDA, are "more volatile than larger borrowers": 35 percent of that cohort have grown earnings by more than a quarter since their loan closed, against 28 percent in the core middle market, and 18 percent have lost more than a quarter of their earnings, "more than double the 8% observed in the core middle market" (same release).
Read those two pairs together rather than separately. The smaller cohort is not simply doing worse. It is doing both better and worse, and the distribution is what matters once a covenant is written against it, because a test is breached by the bad tail and is not rewarded for the good one. A borrower whose earnings swing a quarter in either direction has a materially higher chance of meeting the earnings shock in the table above than one whose earnings move half as much, whatever its average outcome.
The default data sits where that reasoning predicts. Across the whole market, defaults rose to 1.5 percent of outstanding loan principal from 0.8 percent the quarter before, which the release calls nearly doubling, and 3.7 percent by count from 2.5 percent. Among borrowers below 100 million dollars of EBITDA, "the bulk of the direct lending market", they ran at 3.9 percent size-weighted and 5.0 percent by count, from 3.0 and 3.6 percent in the second quarter, and the release states that borrowers above 100 million dollars "remain the least impacted" (same release). Seven percent of all loans are marked below 90 percent of par, which the release calls more than double the historical average. The size divide in those default and valuation figures is argued in full in our September piece on the same DataBank series and is not re-argued here.
The release also closes off the reading that this is one industry's problem. By sector it says "stress remains concentrated rather than widespread", with healthcare highest at 7.1 percent by count and consumer next at 4.6 percent. The pattern that holds across the data is scale, not sector: the release's own summary is that "default risk in private credit remains closely associated with borrower scale".
What have lenders already done about it?
Lent less against the same earnings. The release reports that "lenders are responding by adjusting underwriting discipline", and the adjustment shows up in leverage at origination rather than in price: 2025 and 2026 vintages closed at a median of 4.3 and 3.8 times respectively, against 4.6 times for "2021 and prior loans", and through the first nine months of 2026 leverage at underwriting for small borrower deals "came in a full turn below those of the core middle market" (same release).
That is the part of this release an owner can act on, and in our judgement it is more informative than the default rate. A default rate tells a borrower what happened to other people's loans. A full turn of leverage withheld at underwriting tells a borrower what its own next loan will look like, and it is a decision the lender has already made rather than one it is considering. On earnings of 15 million dollars, a turn is 15 million dollars of debt that a borrower in this cohort will not be offered on terms a borrower a size band up would get.
Note which lever moved. Lenders facing a cohort with wide earnings swings did not principally reprice the risk, they reduced the exposure. That is the orthodox response to variance rather than to level: a higher coupon compensates a lender for a loss it expects on average, while less debt is what protects it from an outcome it cannot predict. The release's framing supports that reading, tying the tightening to "higher observed earnings volatility" rather than to a weaker outlook for the cohort.
“Owners spend their attention on the rate because it is the number that moves every day, and because it is the only one of the two that somebody will sell them protection against. The test in the credit agreement is not primarily a rate test. It is an earnings test with a rate input, and on this data the earnings input is about three times the lever. The practical consequence is that the work which protects a covenant is the unglamorous work inside the business: the contract that renews, the customer that is not a fifth of revenue, the margin that holds when one input price moves. A hedge is worth having. It addresses the smaller half of the problem.”
What should an owner do before the next renewal?
Model the covenant against the earnings case first, in our judgement, and at a shock of the size this data says is ordinary rather than remote. Eighteen percent of borrowers in the 10 to 20 million dollar band lost more than a quarter of their earnings; that is close to one in five, which makes a 25 percent downside a planning case rather than a stress case for a business of that size. The question to put to a model is not whether the business survives it but whether the credit agreement does, and at what headroom.
Three consequences follow for a borrower preparing to refinance or amend. First, the evidence a lender will find persuasive is evidence about earnings stability rather than about growth: the length and renewal history of contracts, concentration in the customer base, how margin behaved the last time an input moved. A borrower that can show a narrow earnings distribution is arguing against the exact number this release puts on its cohort. Second, the structure to negotiate at closing is headroom in the test and the right to cure, not a lower coupon, because the coupon is the smaller lever on the thing that bites. Third, timing favours borrowers who go early: leverage at underwriting has tightened in each of the last two vintages on this data, so a borrower waiting for a better coupon is waiting inside a trend that is moving against the amount it can raise.
The one genuinely reassuring figure in the release deserves to be read with the rest. Most borrowers are at or near their best ever revenue and earnings, and two thirds grew earnings over the year. Underperformance, where the release finds it, is "more common in EBITDA and enterprise value than in revenue, pointing to margin and valuation multiple pressure rather than weaker demand". A business whose top line is intact and whose margin is thinning is the specific profile that the earnings leg of that sensitivity table is built to catch, and it is not a profile that looks distressed from the inside.
What would change this picture?
A second quarter of falling coverage while earnings keep growing, which would make this a direction rather than a print. Coverage slipped this quarter against median earnings growth of 8.1 percent, so the mechanism was cost rather than performance. If the next edition shows the median below 1.68 times with earnings still up, the cost side is doing the work on its own and the sub-1.0 share becomes the number to watch rather than the median.
On the other side, the leverage figures are the fastest signal that lenders have changed their minds. Origination leverage tightened across the 2025 and 2026 vintages on this data. If a 2027 vintage prints back above 4.3 times for small borrowers, lenders will have decided the earnings variance in that cohort was a cycle rather than a feature, and the full turn now withheld will start coming back. If it tightens again, the gap between what a smaller borrower can raise and what a larger one can raise widens for a third consecutive vintage, and that gap, rather than any default rate, is what an owner of a business in this size band will actually meet at the next renewal.
As of 20 November 2026
Sources: Houlihan Lokey, "Houlihan Lokey Finds Size-Weighted Defaults in Private Credit Nearly Doubled in the Third Quarter, Reaching a Near-Term High With Smaller Borrowers the Most Impacted", issued through Business Wire with a Los Angeles dateline of 19 November 2026, reporting the firm's Q3 2026 Private Credit DataBank Market Trends and Insights report, which the release states is available to clients of the firm. businesswire.com and hl.com both refused an automated read, so the release was read in full from its verbatim syndication carrying the Business Wire dateline, and every figure below is taken from that text. For median borrower interest coverage falling to 1.68 times in September 2026 from 1.73 times in June, described as the first decline since mid-2025, and the share of borrowers with coverage below 1.0 times rising to 15.1 percent from 13.9 percent; the statement that interest coverage sensitivities show coverage is more dependent on earnings than rates, with a 100 basis point rise in base rates lowering median coverage to 1.55 times and lifting the sub-1.0 times share to 18.0 percent, and a 25 percent decline in earnings at current rates cutting the median to 1.26 times and doubling the sub-1.0 times share to 30.0 percent; median revenue up 5.9 percent and median EBITDA up 8.1 percent year over year, with 70 percent of borrowers growing revenue, 66 percent growing EBITDA, and 85 percent at or within 10 percent of peak revenue, and the finding that underperformance is more common in EBITDA and enterprise value than in revenue, pointing to margin and valuation multiple pressure rather than weaker demand; market-wide defaults rising to 1.5 percent of outstanding loan principal from 0.8 percent in the prior quarter and to 3.7 percent by count from 2.5 percent, on the DataBank's most fundamental definition of default, which captures technical as well as payment defaults; defaults among borrowers with less than 100 million dollars of EBITDA, described as the bulk of the direct lending market, at 3.9 percent size-weighted and 5.0 percent by count against 3.0 and 3.6 percent in the second quarter, the statement that borrowers above 100 million dollars remain the least impacted, and the summary that default risk in private credit remains closely associated with borrower scale; 7 percent of all loans valued below 90 percent of par, described as more than double the historical average; the earnings volatility of lower middle market borrowers with between 10 and 20 million dollars of EBITDA, with 35 percent growing EBITDA by more than 25 percent since origination against 28 percent in the core middle market, and 18 percent having lost more than 25 percent of EBITDA against 8 percent in the core middle market; the statement that lenders are responding by adjusting underwriting discipline, with 2025 and 2026 vintages closing at median leverage of 4.3 and 3.8 times against 4.6 times for 2021 and prior loans, and leverage at underwriting for small borrower deals through the first nine months of 2026 coming in a full turn below the core middle market; the sector statement that stress remains concentrated rather than widespread, with healthcare highest at 7.1 percent by count and consumer at 4.6 percent by count; and the description of the DataBank as carrying observations back to 2017, refreshed quarterly, with more than 300 standardized data fields on each loan, spanning the full direct lending market from borrowers with less than 20 million dollars of EBITDA to those with more than 100 million. The release contradicts itself on the cohort behind the 11 percent below-par figure, attributing it in a bullet to borrowers with EBITDA below 20 million dollars and in the body to borrowers with 10 to 20 million dollars of EBITDA; that figure is not used in this article, and the band definitions used here are the body's. The reading that the earnings leg rather than the rate leg decides a smaller borrower's covenant, the arithmetic comparison between the two shock rows, the point that a two-sided earnings distribution is penalised by a covenant that is breached by the bad tail and not rewarded for the good one, the observation that lenders answered variance by reducing exposure rather than by repricing it, and every consequence drawn for an owner negotiating, amending or refinancing debt are our judgement and are labelled as such in the copy. The companion articles cited for the size divide in defaults and marks, for the definitions of the coverage tests, for the forward curve and for hedging instruments carry their own sources.
Corrections: factual errors are corrected on the page and the correction dated. Write to admin@kadenwoodcapital.com.
This position sits within our debt advisory practice.

