What did the recovery data actually show?
That every tier of the debt stack now gets back less when a borrower fails. Average recoveries on rated US corporate debt fell from 77 percent to 61 percent for first-lien lenders between the 2008 to 2022 period and 2023 to the first quarter of 2026, on S&P Global Ratings data charted by Apollo's chief economist on 24 August 2026.
The rest of the stack moved the same way. Second- and third-lien recoveries fell from 31 percent to 26 percent across the same two periods, and unsecured recoveries fell from 29 percent to 9 percent, on the same chart.
The sizes of those moves are worth stating in plain terms, and this is our arithmetic on the source's figures. The first-lien decline is 16 percentage points, which removes roughly a fifth of the recovery a senior lender used to count on. The unsecured decline is 20 points, which leaves that class with under a third of what it used to recover. The gap between the top and bottom of the stack has widened from 48 points to 52.
The comparison periods matter. The first covers fifteen years including the financial crisis and the pandemic; the second covers a little over three years in which the headline default rate has been low. Recoveries fell in a period that, on the default gauge, looked calm.
| Debt class | 2008 to 2022 | 2023 to Q1 2026 | Change |
|---|---|---|---|
| First-lien | 77% | 61% | Down 16 points, about a fifth of the prior recovery |
| Second- and third-lien | 31% | 26% | Down 5 points |
| Unsecured | 29% | 9% | Down 20 points, to under a third of the prior recovery |
| Gap, first-lien over unsecured | 48 points | 52 points | Widened by 4 points |
| Direct lending index, realized losses from defaults | 1.0% long-term average | About 0.5% | Stable at one-half the long-term average, at 30 June 2026 |
Why is a lender recovering less in a low-default market?
Because of two changes in what defaults look like when they arrive, according to the source. Apollo attributes the decline to out-of-court distressed exchanges that reshuffled priority without real deleveraging, and to a growing share of asset-light software borrowers that left creditors with little tangible collateral to seize.
The first cause is a mechanism. A distressed exchange, the liability management exercise that a companion article on this site tracks in the monthly default reports, typically resolves stress by moving some lenders up the priority ladder and others down, extending maturities, or swapping debt at a discount. The company survives. The debt load often does not fall much. When the same borrower fails a second time, there is less value left and more claims ranked ahead of whoever was pushed down. That second failure is where the lower recovery is recorded.
The second cause is a balance sheet. A borrower whose value sits in code, contracts and staff rather than plant, inventory and property offers a lender the right to seize things that are worth little without the people who built them. A first lien over an asset-light business is senior to everything and secured by not much. The chart shows what that is worth when it is tested.
Neither cause is about interest rates or the economy, and that is what makes the series useful. Recoveries fell for structural reasons, and structural reasons do not reverse when the cycle turns.
Does the same decline show up in private middle-market lending?
Not yet, on the one broad index that reports it. Cliffwater's Direct Lending Index, which covers more than 23,000 directly originated US middle-market loans totalling $553 billion, reported realized losses from defaults stable at one-half of their 1.0 percent long-term average as at 30 June 2026, in its mid-year release of 19 August 2026.
The same release put the index return at 3.0 percent for the first six months of 2026 and 7.7 percent for the trailing year, with almost 22 years of returns averaging 9.5 percent and one negative year, 2008. First-quarter price markdowns of 3 percent in software loans, which dampened returns, were followed by second-quarter markdowns of under 1 percent. Non-interest cash flow from maturities and prepayments stayed above 6 percent of the index per quarter.
So the two prints describe different populations and, for now, different outcomes: a rated, largely syndicated universe where recoveries have fallen at every tier, and a directly originated middle-market universe where realized losses are half their long-run norm. The software markdown that Cliffwater reports and the software borrowers Apollo names are the same exposure viewed from two sides, and it is the point at which the two series are most likely to converge.
A companion article on this site sets out how contested the private-credit recovery evidence is, with datasets that range from around fifty cents to a rated cohort well above that. The Apollo chart does not settle that dispute. It shows which direction the reference market is moving while it goes on.
What does a recovery chart change for a borrower who is not in it?
The price of collateral, and this is our reading rather than Apollo's or Cliffwater's, neither of which writes for a middle-market owner. A company of the size most owners here run is not rated and not in either index. Its lender's underwriting model is still built on assumptions about recovery, and those assumptions have just been revised down by the reference series every credit committee reads.
The first consequence is a re-rating of tangible assets. When first-lien recoveries in the rated market fall by a fifth and the source names asset-light borrowers as one of two causes, a lender pays more attention to what it could actually seize. An owner with plant, property, equipment or receivables that can be independently valued is offering something the market has just repriced upward. An owner whose value sits in people and relationships is asking a lender to extend senior debt against a recovery estimate that the reference data now puts nearer 61 percent than 77 percent, and the lender will price, structure or size that loan accordingly.
The second consequence is that the order of the stack matters more than it did. With a 52-point gap between first-lien and unsecured recoveries, a lender's willingness to be anything other than first-lien senior secured shrinks, and any junior or unsecured tranche in a middle-market structure gets priced for a nine-cent outcome. Owners who want a subordinated layer should expect it to cost what that recovery implies.
The third is that the distressed exchange has a price for the borrower too. The mechanism that lets a company avoid a payment default by reshuffling its lenders is the same mechanism the source blames for lower recoveries on the second failure. A lender that has read the chart will negotiate the first exchange harder, because it now knows what the second one is worth.
“A first lien over a business with no hard assets is senior to everything and secured by not much. The chart puts a number on that for the first time this cycle, and every lender across the table has already read it.”
What would change this picture?
The Cliffwater loss rate, first. Realized losses at half the long-term average are the private middle market's evidence that it has not followed the rated series down. If that rate rises toward or through 1.0 percent over the next two releases, with the software markdowns as the driver, the two populations are converging and the owner-side reading above hardens.
The composition of the rated series, second. The decline is attributed to distressed exchanges and asset-light borrowers, which means it is concentrated rather than uniform. A later cut that separated hard-asset borrowers from software would show whether an owner with a plant is being tarred with a software recovery, or whether the decline is broad.
The lien gap, third. Fifty-two points between first-lien and unsecured is the widest reading in the chart's two periods. If distressed exchanges keep pushing junior classes down the ladder, the gap widens further and the case for anything but senior secured weakens. If the exchange market cools, the gap should stabilize.
Hold the frame where the sources put it. The recovery figures are averages on rated US corporate debt across two periods, charted by Apollo from S&P Global Ratings data on 24 August 2026; the loss and return figures are Cliffwater's, on its direct lending index at 30 June 2026. Nothing here is a statement about any individual borrower, lender or fund, and the collateral reading in the previous section is ours.
As of August 2026
Sources: Apollo Global Management, The Daily Spark by Chief Economist Torsten Slok, 'Recoveries Are Falling as Distressed Exchanges and Software Borrowers Leave Less Behind', 24 August 2026, with the chart sourced to S&P Global Ratings: for average recoveries on first-lien debt of 77 percent in 2008 to 2022 against 61 percent in 2023 to the first quarter of 2026; for second- and third-lien recoveries of 31 percent against 26 percent; for unsecured recoveries of 29 percent against 9 percent; and for the stated causes, that out-of-court distressed exchanges reshuffled priority without real deleveraging and that a growing share of asset-light software borrowers left creditors with little tangible collateral to seize. Cliffwater, 'Private Credit Shows Resilience in Cliffwater Direct Lending Index Data', published 19 August 2026 via PR Newswire, measured at 30 June 2026: for the Cliffwater Direct Lending Index returning 3.0 percent for the first six months of 2026 and 7.7 percent for the trailing year; for almost 22 years of returns averaging 9.5 percent with one negative year, 2008; for realized losses from defaults remaining stable at one-half of their 1.0 percent long-term average; for first-quarter price markdowns of 3 percent in software loans and second-quarter markdowns of less than 1 percent; for non-interest cash flow from private loans remaining above 6 percent per quarter; and for the index covering over 23,000 directly originated US middle-market loan holdings totalling $553 billion in assets. The 16-point, 5-point and 20-point declines, the descriptions of those declines as about a fifth and under a third of the prior recovery, the 48-point and 52-point first-lien-to-unsecured gaps and their widening, the description of the two populations as different, and the entire collateral and middle-market reading are our own arithmetic and judgement and carry no Apollo, S&P Global Ratings or Cliffwater figure. Companion articles on this site cover the contested private-credit recovery evidence, the July 2026 default and distressed-exchange gauges, what a covenant breach does and does not mean, and what happens when lenders take the keys.