85 Cents of Every Investment-Grade Dollar Is the Government's Price. Credit Is Only the Rest.
Start with the 2 layers of every corporate bond. The bottom layer is the risk-free rate: what the US government pays to borrow for the same number of years. No treasurer can negotiate it. The top layer is the spread, which is the extra yield a lender demands because the company might not repay. Only that top layer is about the company.
When the government's own 10yr borrowing rate is 5.10% and an investment-grade company pays 0.92 points of spread on top, the company's all-in cost is 6.02%. About 85 cents of every dollar of that interest bill is a judgment about the US government. The company's own credit risk is the remaining 15 cents. That split is normally a footnote. This month it is the story, because the supply of the bottom layer changed.
The price of long-dated money rose while the supply of long-dated bonds fell
Take the term premium first. That is the extra yield a buyer demands for lending for 10 years instead of rolling short bills - the price of tying money up. The New York Fed estimates it monthly and daily; its 10yr measure printed 0.9864% on 13 November.
Set that against 2 other dates. On 5 November the same measure printed 0.8693%. On 6 October it printed 1.0844%. So October took about a fifth of a percentage point out of the price of duration, and November has put a little over a tenth of a point back, nearly all of it after the 5th.
Of the 251 sessions in the past year, 19 closed higher. That ranks this print around the 92nd percentile. In plain terms: long-dated money is expensive, though not at its most expensive point of the year.
Now the supply side. Over the 90 days to 16 November the Treasury settled 814.8bn dollars of notes and bonds at auction. In the prior 90 days it settled 1,234.2bn. That is a fall of 419.4bn, or 34%. Notes and bonds were 12.4% of everything the Treasury sold in the recent window, against 15.3% before. It funded the difference at the bill line, which is debt that reprices in weeks rather than decades.
If you take duration out of the market, the price of duration should fall. It rose 11.7 hundredths of a percentage point instead. That contradiction is the piece.
The mechanism: a government that intends to legislate does not want to lock in a 30yr coupon
Here is how the political fact reaches a price. On 3 November the Democrats took the Senate and were projected to take the House. That ended the divided Congress this series had been priced against.
A government with an open fiscal channel has a specific reason to fund short. A 13-week bill reprices almost immediately, so the sovereign is not stuck with a coupon set under a policy path it is about to change. A 30yr bond is a 3-decade bet on inflation and on the political settlement that produces it. A government that plans to legislate is the least willing buyer of its own long-dated paper.
Duration supply does not vanish when the sovereign steps back. It migrates. Every unit of long-dated exposure in the system now has to be manufactured by someone else: a company issuing a 10yr bond, a household taking a 30yr mortgage, a bank holding long assets against deposits. Scarcity of the government's duration does not make the private sector's duration cheap, because a treasurer cannot issue Treasury bonds. They are not the same product.
So the scarcity shows up as a higher price, not a lower one. The term premium is what clears the duration nobody else is supplying. And because the corporate curve is priced off the government curve, that premium lands on a treasurer's desk whether the treasurer wants it or not.
In investment grade, the company's credit is 92 hundredths of a point out of 602
Investment-grade spreads printed 0.92 points on 17 November, against a 10yr government close of 5.10% the same day: a 6.02% all-in borrowing cost.
High yield is the other half of the comparison. That index printed 3.43 points, so the all-in is 8.53%. The government's share drops to about 60% and credit's share rises to 40%. The extra spread for the riskier borrower is 343 hundredths of a point against 92 - nearly 4 times as much - for a default risk that nobody seriously thinks is 4 times higher.
This is where the AI buildout argument actually lives. A datacenter gets underwritten on an all-in cost of money and a depreciation schedule: 15 years of useful life, financed at 6.02% if the issuer is investment grade. Of that 6.02%, 5.10 points are a judgment about the US government's duration. Not about the building, the chips, or the tenant. When the price of duration moves 11.7 hundredths of a point in 6 sessions, every project in the pipeline gets a slightly worse return, and no amount of demand for computing changes that arithmetic.
Read the 2 all-in numbers side by side and the asymmetry matters more than the levels. A 10 hundredth-of-a-point move in the term premium adds 10 to both numbers, which is 1.7% of the investment-grade total and 1.2% of the high-yield total. A 10 hundredth-of-a-point move in the investment-grade spread adds 10 to one number only. The government leg is the bigger, the faster-moving, and the less diversifiable of the 2.
The quality gap is the one instrument where credit trades on its own
The quality gap is high-yield spread minus investment-grade spread. It measures what the market charges for the difference between a borrower likely to repay and one that might not. It printed 251 hundredths of a point on 17 November.
Trace the path. On 5 November it printed 255, the highest of the year. On 12 November it was 231. On 16 November, 244. On 17 November, 251. So it made a 1-year high 2 sessions after the vote, gave back 24 hundredths of a point, and has since taken 20 of those 24 back - about 83%.
That 20-point move over 3 sessions is the 27th largest in 781 such moves since November 2023, around the 97th percentile. Large, but 26 larger moves exist, mostly in April 2025 and August 2024.
The level matters more than the move. At 251, exactly 2 of the past 264 sessions closed higher: 253 on 30 March and 255 on 5 November. The 1-year median is 205. So this is a high zone on a 1-year view.
Then look at the 3-year record, which is the honest correction. The maximum in that file is 341 hundredths of a point, printed on 7 April 2025 during the tariff shock, and 69 of 784 sessions have printed at or above 251. On the longer record this is expensive, not extreme. Anyone selling it as a crisis call is overselling it.
The 2 legs also moved at very different speeds. High yield went from 3.20 to 3.43 points, 23 hundredths of a point or 7.2%. Investment grade went from 0.89 to 0.92, 3 hundredths of a point or 3.4%. The bottom of the stack moved nearly 8 times as far in absolute terms.
That is what makes the quality gap the political instrument rather than the general stress gauge. The gauge that ran yesterday moved mostly on its volatility leg, which is a statement about the price of insurance. The quality gap is a statement about which borrower the market is repricing, and it is the only line on the credit tape that changed direction twice in 8 sessions.
What argues against this
Two measurements cut the other way, and both are in the data.
Volatility is one session ahead of the spreads. The VIX closed at 24.56 on 18 November from 25.46, down 0.90, and the rates volatility index closed at 98.15 from 101.88. If the insurance market is right, cash credit should follow it lower. The next spread prints should land near 244, the 16 November gap, not above 251. A gap that holds above 244 while volatility falls is a genuine repricing of credit. A gap that slides back to 231 is a flow that briefly ran.
The percentile is a 1-year statement, not a history. On the 3-year file this is a high zone, not an extreme, and the piece should be read that way.
The third objection is the least comfortable. The 20 hundredth-of-a-point give-back happened during a broad risk move, so part of it is general market direction rather than politics. The defensible claim is narrower than the headline: the quality gap is where the political variable would have to appear first, and it has now failed twice to hold a retracement. That is evidence about the direction of pressure, not proof of its cause.
The instrument and the falsifier
The instrument is the quality gap at 251 hundredths of a point, around the 99th percentile of the past year against a median of 205. The second instrument is the investment-grade all-in cost of 6.02%, of which 5.10 points are the government leg and 85% of the total.
The claim is wrong if any of 4 things happen.
Below 231 with high yield at or under 3.20%: the give-back was flow, the retracement resumes, and the political read goes with it.
Investment grade above 0.94% and high yield above 3.48% on the same date: both legs at 1-year highs at once is a repricing of credit, not of quality, and the "85 cents is the government" framing stops being the point.
Investment-grade spreads holding at or above 0.92% for 3 consecutive prints while the 10yr falls to 4.90% or lower: that means credit, not duration, is the binding constraint, and the sovereign-share argument is describing the wrong leg.
The term premium back below 0.8693% while the note-and-bond share stays at or under 12.4%: that makes the 11.7-point rise a positioning event rather than the price of withdrawn duration, and the mechanism section is then a story about nothing.
The confirming case is a gap that holds above 244 while the VIX stays under 25.46, with the term premium above 0.8693%.
Notes on the numbers
Term premium: the ACM 10yr estimate published by the Federal Reserve Bank of New York, last print 13 November at 0.9864%. That series runs a few days behind the cash curve, so every use of it here is dated to the print.
Credit: the ICE BofA US High Yield and US Corporate option-adjusted spreads - the spread after adjusting for the bond's option features - both as of 17 November, 3.43% and 0.92%. The all-in numbers use the 17 November government 10yr close of 5.10%, the same date as the spreads, not the 18 November close of 5.07%.
Curve, 18 November close: 2yr 4.80, 3yr 4.84, 5yr 4.89, 7yr 4.96, 10yr 5.07, 20yr 5.44, 30yr 5.41. The 2yr-to-10yr slope is 27 hundredths of a point, unchanged from 13 November, so the whole 8-point fall in the 10yr was a parallel move.
Issuance: gross auction settlements from the Treasury's Fiscal Data service, over 2 rolling 90-day windows ending 16 November and 17 August. Bills are every tenor of 52 weeks or less. These are gross settlements rather than net borrowing, so they describe what the market had to absorb, not the deficit.
Left out deliberately: the 30yr mortgage series, last printed 22 October, and the quarterly median home price, last printed for 2026-Q3. Housing is the obvious third leg of this argument and both series are 4 weeks or more behind the claim, so they are excluded rather than dated into an argument they cannot carry.