The model passes through about 56% of the 60-day median 10-year TIPS real yield, with the result floored at 0%:

fixed rate ≈ max(0, 0.556 × 60-day median 10Y real yield)

-

Why 0.556, and why a floor

Since 2018, Treasury's fixed rate has closely tracked about 56% of the 10-year TIPS real yield. (Earlier years followed different patterns, so we only fit to resets from 2018 on.) We floor the estimate at 0% because Treasury has never set a negative fixed rate: a 0% I‑Bond already beats a 0% TIPS, since it can't lose value and defers federal tax.

About the 90% range

The range is a 90% interval that combines two sources of error: how well the model has fit past resets (wider when recent yields have been volatile), and how far real yields could drift between now and the announcement (wider the further out we are). Both are estimated from 2003-present data. It does not price in Treasury changing its approach.

• The range already includes yield drift. It widens with the horizon to the reset and with recent volatility, so it accounts for real yields moving between now and the announcement - but only within the range of moves seen since 2003.
• Treasury has discretion, and has changed its approach before. There is no published formula; this is a fitted relationship, and Treasury's rate-setting behavior has shifted at least three times since 1998. If it shifts again, these numbers go stale.
• Small sample. Treasury's rate-setting has shifted over the years, so its current approach only covers a handful of resets. There just isn't much data behind this yet.
• Not investment advice. This is a statistical estimate from public Treasury data, not a recommendation to buy, hold, or sell.