The 2026 yield curve as a constraint on fixed-income positioning

A US Treasury par yield curve does not look interesting at a glance. On 31 July 2026 the curve runs from 3.78 percent at 1M, through 3.83 percent at 3M, 3.98 percent at 6M, 4.08 percent at 1Y, 4.28 percent at 2Y, 4.34 percent at 3Y, 4.45 percent at 5Y, 4.59 percent at 7Y, 4.75 percent at 10Y, 5.28 percent at 20Y, and 5.27 percent at 30Y (US Treasury, Daily Treasury Par Yield Curve Rates, 31 July 2026). The Federal Reserve H.15 Selected Interest Rates release on the same day shows the constant-maturity column within 4 basis points on every tenor.

The constraint, for a family office book, is not the level. It is the spread between 1M and 30Y, 149 basis points on the print, the widest in the current cycle. That is the line the desk watches when it is asked to write a fixed-income policy that holds through the next twelve months without being re-signed every time the Federal Reserve moves.

The Federal Reserve Monetary Policy Report submitted to Congress on 10 July 2026 is explicit on the asymmetry. The 2-year nominal Treasury yield is up about 60 basis points since the start of the year. The 10-year is up around 35 basis points on net. The largest moves have been at shorter maturities, a pattern the report attributes to a higher expected path for the federal funds rate after the Middle East conflict and to a firmer read on US labour market stability.

The short end had already done most of its travel before January. The 1M par yield closed June 2024 at 5.47 percent, fell to 4.28 percent a year later, dropped to 3.72 percent at end-January 2026, and sat at 3.78 percent on 31 July 2026 (US Treasury). That is roughly 170 basis points of travel inside the window in which most family offices signed their last fixed-income policy.

A 170 basis point move on the short end inside thirty months is not a once-a-decade event. It is the operating environment. A policy written as a target yield at the front end, rather than as a band around the spread to the long end, is one Fed meeting away from needing to be re-signed.

The full curve carries a 3.78 percent 1M to 4.08 percent 1Y base, then a 4.28 percent 2Y, a 4.45 percent 5Y, a 4.75 percent 10Y and a 5.27 percent 30Y. Cash held in 1Y Treasuries at the back end of the curve is paying 30 basis points over the 1M. Five-year paper is paying 67 basis points over 1Y. Ten-year paper is paying 97 basis points over 2Y, and the 30Y is paying 52 basis points over the 10Y.

Those are the lines the desk uses on a mandate to size the cost of staying short. A book that sits in 1M T-bills for liquidity and reaches for 5Y to 10Y on the operating reserve is being paid roughly 50 to 100 basis points of carry for the duration it adds. A book that reaches for the 20Y to 30Y for the long reserve is being paid another 50 basis points on top of that. The whole trade, sized honestly, is a few hundred basis points of carry at most.

Most family-office fixed-income policies are written as target allocations, with or without an allowed drift. A 60/40 equity-to-fixed-income policy with no band is read by the desk as a policy that gets re-signed when the S&P 500 moves by more than 10 percent in a quarter. A policy that is written as a band, say 50 to 70 percent equity, 30 to 50 percent fixed income, with a written rebalance trigger at the band edge, can hold through the same quarter without a re-signature.

The same logic applies inside fixed income. A policy that says hold 30 to 50 percent of fixed income at the 1Y to 5Y point of the curve, 30 to 50 percent at 5Y to 10Y, 0 to 20 percent at 10Y to 30Y, and the rest in 0Y to 1Y, is a policy that holds through a 60 basis point front-end move. A policy that says hold X percent at the 10Y point is a policy that gets re-signed every quarter.

The 31 July 2026 curve makes the point. A 1Y to 5Y band holds a carry between 4.08 and 4.45 percent. A 5Y to 10Y band holds a carry between 4.45 and 4.75 percent. The trigger for moving between bands is not a Fed move. The trigger is the carry at the band edge relative to the carry at the next band up. As long as the 5Y to 10Y band pays a positive carry over the 1Y to 5Y band, the policy says to be there.

Three rules cover most of the work. The first is the band structure by tenor, with a written rebalance rule at the band edge. The second is a duration budget set in standard deviations of the trailing two-year band on the 10Y, not in dollars of yield. The 10Y has held in a 4.24 to 4.75 percent band since mid-2025, a 51 basis point range; setting the trigger at one standard deviation gives a band the policy can hold through without a re-signature.

The third rule is the link from the curve to the rest of the book. The fixed-income band is not a free-standing allocation. It is sized against the liquidity need for the next twelve months, the FX exposure of the office, and the unfunded commitments in the private book. On most retained mandates the fixed-income band is 30 to 50 percent of the book, the equity band is 30 to 50 percent, and the alternatives and gold band is 0 to 20 percent, with the rest in cash and 0Y to 1Y paper.

That is the whole policy. It does not need a re-signature when the 1M moves 30 basis points, or when the 10Y re-tests 4.8 percent, or when the Fed holds or cuts. It needs a re-signature only when the principal changes the season the office is funding for.

The 31 July 2026 print is the curve the desk will use for the next quarter. Short end 3.78 percent, long end 5.27 percent, 1M to 30Y spread 149 basis points. The Federal Reserve Monetary Policy Report, the Treasury yield-curve page, and the H.15 release all agree on the print within a few basis points. Three independent primary sources, one set of figures, no extrapolation required.

The reading for the family office book is the one we have carried through the year. Stay in the bands. Treat the spread as the policy lever, not the level. Re-read the bands every quarter, and re-sign only when the season changes. The curve will keep moving. The policy that survives the cycle is the policy that does not need to be re-signed every time the curve does.