The High-Yield Average Is Hiding Its Tail
There are two numbers in the American corporate bond market right now, and they do not agree with each other. One of them is an average. The other is the part of the market the average is built to smooth over.
The first number: the option-adjusted spread on the ICE BofA US High Yield Index, the benchmark for the entire junk-bond universe, was 2.71 percent on Aug. 13, 2026, according to the series published by the Federal Reserve Bank of St. Louis. That is roughly 271 basis points of compensation over Treasurys for owning the debt of companies the rating agencies do not consider investment grade. It is a number that reads like nothing is wrong.
The second number: the option-adjusted spread on the ICE BofA CCC and Lower US High Yield Index, the bottom rung of the same universe, was 10.24 percent on the same date, on the same St. Louis Fed series. That is a number that reads like something is wrong.
Subtract one from the other and you get 8.64 percentage points. That figure is not published anywhere; it is arithmetic performed on two sourced series, both observed Aug. 13, 2026, and it should be read as arithmetic rather than as a quoted market level. But it is the cleanest single summary of what has happened to corporate credit this year.
The index option-adjusted spread is built from each constituent bond's own spread, weighted by market capitalization, across everything in the index. That construction is a feature when you want one number for a market, and a limitation when the market stops being homogeneous. A cohort that is small by market value can widen dramatically without dragging the headline number with it, because the headline is being held down by the much larger, much better-rated bulk of the index. The index is not lying. It is answering a different question than the one investors think they are asking.
Walk the gap forward through 2026 using the two St. Louis Fed series and the shape is unmistakable. On Jan. 2, 2026, CCC and lower stood at 8.88 percent and BB at 1.71 percent, a difference of 7.17 points. On March 2 the two were 9.45 and 1.82, a difference of 7.63 points. On June 1 they were 9.46 and 1.62, or 7.84 points. On July 1, 9.68 and 1.63, or 8.05 points. On July 31, 10.34 and 1.73, or 8.61 points. And on Aug. 13, 10.24 and 1.60, or 8.64 points.
The widening has come from both directions at once, which is the part that matters. The BB cohort, the highest-quality tier of high yield, has tightened over the year: 1.71 percent on Jan. 2, 2026 falling to 1.60 percent on Aug. 13, 2026. The CCC cohort has gone the other way, from 8.88 percent to 10.24 percent across the same span. Investors have not become more cautious about high-yield credit. They have become more cautious about a specific slice of it while becoming less cautious about the rest.
The BB series reached 1.56 percent on June 17, 2026, matching the lowest reading anywhere in the three years of history the St. Louis Fed now carries for that series — a floor it first set on Feb. 18, 2026 and touched again on June 22. That floor sits inside the same stretch in which the CCC series was grinding higher.
The CCC series has not made a new extreme, and it is worth being precise about that. Its 2026 high was 10.34 percent on July 31. Its highest reading in the St. Louis Fed's three-year window was 11.37 percent on April 7, 2025, during the tariff-driven repricing that spring. The current level is elevated relative to this year, not relative to the worst of the last three.
A caveat about history that changed the reporting on this story. The St. Louis Fed's ICE BofA spread series carry a note stating that starting in April 2026 they will include only three years of observations, and the files now begin on Aug. 15, 2023. It is no longer possible to establish the long-run extremes of these series from the Fed's own database. Anyone citing a 2007 low or a 2008 high for high-yield spreads is citing a secondary compilation, and should say so.
Trading Economics, which maintains a longer reconstruction of the series and cites Federal Reserve data as its source, puts the record low for the broad high-yield spread at 2.41 in June 2007 and the record high at 21.82 in December 2008, across a history it runs back to 1996. That is a secondary source and is presented here as one.
Which brings up a framing worth correcting. The broad high-yield spread is often described as sitting at multi-decade lows. On the Fed's own three-year file, it is not even at a three-year low: the series printed 2.59 percent on Jan. 22, 2025, below the 2.71 percent observed on Aug. 13, 2026. Over that three-year window the series has ranged from that 2.59 percent trough to 4.61 percent on April 7, 2025. The honest description is that the broad index sits near the bottom of its recent range, not at the bottom of it.
Investment-grade credit shows the same placidity with even less drama. The ICE BofA US Corporate Index option-adjusted spread was 0.79 percent on Aug. 13, 2026, per the St. Louis Fed, against a low of 0.73 percent reached at several points earlier this year. Nothing in the high-grade market is signaling stress.
Outside observers noticed the divergence well before it reached its current width. Alpinum Investment Management wrote on June 17, 2026 that "Consequently, high yield investors now require about 6.4 percentage points of extra yield to own high-risk CCC-rated bonds over BB notes, marking a 14-month high." The firm added, in the paragraph that follows, that "This market dispersion intensified after a synchronized spike in March 2026."
That 6.4 figure should not be stacked against the 8.64 computed above as though the two were the same measurement. On the day Alpinum published, the Fed's CCC series stood at 9.39 percent and its BB series at 1.56 percent, a difference of 7.83 points, not 6.4. Alpinum does not identify its index or vendor in the piece, and is evidently measuring on a different basis. What carries across is the direction and the fact that a manager flagged it in June; the levels do not carry across at all.
Moody's made the structural version of the argument in April. In a piece dated April 28, 2026 on US corporate default risk, the firm wrote: "These improving averages, however, mask dispersion." It continued: "High yield issuers are larger, more diversified, and less vulnerable to near-term refinancing pressure than the broader universe of public companies." That is a caution about reading the high-yield index as a proxy for corporate America generally, and it applies with equal force to reading the high-yield index as a proxy for high-yield borrowers generally.
The same Moody's piece contains a figure that cuts against any tidy bearish reading, and it belongs in the story. "As of March 2026, the average one year expected probability of default (PD) for all US listed companies stood at 7.9%, down from 9.1% a year earlier, but still elevated by historical standards." Both halves of that sentence are load-bearing. Default risk across listed companies improved year over year, and it is still high by the standards of its own history.
Moody's also reported that for US high-yield companies specifically, "For US high yield companies, PDs declined to 3.2%, reflecting gradual improvement but remaining within the same sideways range that has defined the market since 2023." And on where the risk concentrates: "Smaller, unrated firms, many of which resemble private credit borrowers more closely, continue to exhibit higher and more persistent risk." The tail is not evenly distributed, and a meaningful part of it does not sit in any index at all.
The leveraged loan market shows a parallel split, on a different instrument. Oaktree Capital's Robert O'Leary and Armen Panossian wrote on May 14, 2026 that "There's been no let-up for CCC-rated loans, with spreads widening by over 300 bps so far this year." In the same commentary they wrote that "BB-rated loan spreads have marginally tightened this year, making for a yield of just over 6% – about a quarter of the level of CCCs." Those are loans, not bonds, and the two markets have different investor bases and different documentation. The shape of the divergence is the same one.
None of this has registered anywhere else. The S&P 500 closed Friday, Aug. 14, 2026 at 7,785.76, down 0.17 percent on the day and within a rounding error of its highs. The Cboe Volatility Index closed at 14.63 on Aug. 13, 2026, per the St. Louis Fed's series; the Aug. 14 close had not posted to that series as of Saturday midday and is not reported here. Equity volatility is priced for calm.
The macro backdrop is doing its own thing in the middle. The Bureau of Labor Statistics reported on Aug. 7, 2026 that nonfarm payrolls fell by 23,000 in July and the unemployment rate was 4.1 percent, with May and June revised down by a combined 103,000. The federal funds target range stands at 3.50 to 3.75 percent. A softening labor market is exactly the environment in which the weakest borrowers separate from the strongest, and exactly the environment in which an index average keeps looking fine while doing so.
So what can the headline number actually tell you? It can tell you the cost of capital for the median high-yield borrower, and that cost is low. What it cannot tell you is how many borrowers are nowhere near the median, because the arithmetic that produces it is designed to wash exactly that out. An investor who owns the index owns the tail whether or not the index's spread reflects it.
The next observations for these series are scheduled for Monday, Aug. 17, 2026, per the St. Louis Fed's release calendar. The question they will answer is a narrow one: whether the widest CCC-to-BB gap of 2026 was a July and August event, or the start of something the average has not gotten around to reporting yet.
