[Week 41 of 2026] Yield Signs

Welcome back to Price and Prejudice with a few musings from Week 41 of 2026. Today we talk about what might be driving the recent increase in yields and where they might go from now.

A bond yield is basically a price, and like most prices it can rise for more than one reason at the same time. As always, it helps to think of the yield on a 10-year Treasury as the sum of a few pieces. One piece is the average short-term interest rate that investors expect the Fed to set over the next ten years. Another is compensation for inflation, since the bond pays a fixed number of dollars rather than something that rises with inflation. The last piece, the term premium, is the extra return investors require for holding a long bond rather than rolling over short ones, i.e., for bearing the risk that rates move against them in the meantime. When yields move, the usual explanation names a little bit of everything, and it is hard to tell which piece did the work.

This year the pieces have added up to a lot. The 10-year Treasury yield went from 4.18% at the end of 2025 to 5.24% on October 9, and the 30-year yield is now 5.60% (Treasury). Some people even say that the 10-year could reach 6% for the first time since 2000. In the background, the Fed raised its policy rate to 3.75-4.00% in September, and the meeting minutes show most participants expecting another increase by the end of the year. At the same time, the price of oil is up about 68% since January.

To figure out what's driving these bond yields, it's useful to ask what would have to be true for each explanation to be right (also a popular framework in business strategy). To do this, we can first split the yield into three parts:

10-year yield = expected real short rates + expected inflation + term premium

None of these pieces is observed directly, but there are two standard ways to estimate them. The first way uses inflation-protected Treasuries (TIPS). The TIPS yield measures the real part of the yield, and the gap between the regular yield and the TIPS yield, called “breakeven” inflation, measures the inflation part. The second uses a term structure model, like this one at the New York Fed, which splits the yield into expected short rates and the term premium. It's important to remember that each split is only an estimate, and importantly the two group the pieces move differently in response to different shocks.

So let’s first think about the inflation explanation. If investors expect more inflation (perhaps because oil prices have risen so much) then breakeven inflation should rise. But it has barely moved. Over the year, the 10-year yield rose 106 basis points, of which 98 came from the TIPS yield and only 8 from breakeven inflation, and the market’s measure of expected inflation five to ten years from now also rose by about 8 basis points. Over the last two weeks (September 25 to October 9), breakeven inflation actually fell by a basis point. Whatever is going on, investors do not seem to expect the Fed to let inflation run, which is notable given what oil has done.

Next is the Fed explanation, i.e., investors expect the Fed to keep raising rates. If that is right, the expected-short-rate piece should rise, the 2-year yield (which mostly reflects the next couple of years of Fed policy) should rise, and the yield curve should flatten. For the year as a whole, this fits well. The 2-year yield rose 133 basis points, more than the 10-year, and the New York Fed’s model attributes about 80 of the 10-year’s 106 basis points to higher expected short rates. However, for the last two weeks, it does not fit. The 2-year yield fell slightly, the curve steepened, and the model’s expected-rate piece fell by about 7 basis points.

That raises a follow-up question, which is why investors expect higher short rates this year. One answer is oil. The Fed may be raising rates to keep an oil shock from turning into lasting inflation, which would also explain why breakeven inflation has barely moved. The other answer is AI. Tech companies want to spend enormous amounts on data centers, chips, and power, and when the economy as a whole tries to invest more than it saves, the interest rate that balances the two has to rise, and the Fed ends up following it.

A related AI explanation is supply. Governments are running large deficits, and AI companies have started issuing long-dated bonds as well. A 30-year bond from a hyperscaler carries interest-rate risk just like a 30-year Treasury, so every new long, fixed-rate bond adds to the total amount of interest-rate risk that investors have to hold. Some holders, like central banks and pension funds hedging their liabilities, buy regardless of price. The rest has to be absorbed by investors with limited capital, who need to be paid more to take on more risk, and that extra pay is a higher term premium.

Actually, a few other explanations predict the same thing for the term premium, which makes them hard to separate from supply. If bonds have stopped working as a hedge for stocks, investors will demand more to hold them, and bond and stock returns should start moving together. If mortgage investors and futures traders are selling Treasuries to hedge, as the WSJ has reported, the moves should be sharp and partly reverse. And if the problem is coming from Europe, US yields should move with European yields.

So what does this framing buy us? It shows that the two most common stories can both be right, about different windows. Most of this year’s rise looks like a higher expected path for the Fed’s policy rate, which is consistent with the AI story and a Fed that is fighting an oil shock, while the last two weeks look like a rise in the term premium. And this distinction matters because the Fed plays two different roles in each story behind the rise.

Other Stuff

• Toby Nangle at the FT explains how the roughly $7 trillion agency mortgage-backed securities market can create a “vicious loop” for Treasury yields: when yields jump, mortgage holders sell duration, which pushes yields up further. His read is that this unwind has likely played out.
• Bloomberg reports that long-shot bets, which lose about 98% of the time, dominate trading on Kalshi and Polymarket. Bettors overpaying for long shots, the favorite-longshot bias, is one of the oldest findings in the economics of betting markets.
• Anthropic estimates that about 80% of job tasks, weighted by working time, are exposed to either robots or LLMs. But robots are cost-competitive for just 0.3% of tasks today, and at past rates of price decline it would take about 40 years to reach 10%.