Does the Fed Really Set Interest Rates? Yield Curves, Term Premium, and r-Star Explained

Yields and policy rates change constantly. Specific numbers are anchored to early May 2026 and will move; the framework is descriptive.

Each time the Fed cuts the federal funds rate, a predictable cycle of news stories follows: someone is upset that mortgage rates did not fall, that CD yields fell faster than money markets, or that the 10-year Treasury moved the wrong direction. The reaction makes sense if you believe the Fed sets “the” interest rate. It is not what happens.

The Fed sets one rate directly: a target range for the federal funds rate, currently 3.50 to 3.75 percent as of the April 29, 2026 FOMC meeting. Two administered rates (interest on reserve balances at 3.65 percent, the overnight reverse repo offering rate at 3.50 percent) keep the effective fed funds rate inside that range. Everything else (the 2-year Treasury, the 10-year, mortgages, corporate bond spreads, CD rates, savings yields) is set by markets. The Fed influences those rates, but does not choose them.

This guide walks the actual yield-formation chain: what the Fed controls, how the yield curve decomposes into expected short rates plus a term premium, why r-star matters even though we cannot observe it, why mortgage rates are not just the Fed rate plus a spread, and what all of this implies for DIY investors managing cash, bonds, and mortgages.

What the Fed Actually Controls

The FOMC sets a target range for the federal funds rate, the rate at which depository institutions lend reserves to each other overnight. In an ample-reserves regime (the Fed’s post-2008 framework, after extensive QE), the Fed enforces the target range with two administered rates and two facilities:

IORB (Interest on Reserve Balances): currently 3.65 percent. This is what banks earn on reserves held at the Fed. It acts as the floor of the corridor: no bank lends below IORB if it could just leave money at the Fed and earn it.
ON RRP (Overnight Reverse Repo): currently 3.50 percent. Money market funds and GSEs that cannot earn IORB use this facility instead. It is the floor for them.
Discount Window: a ceiling at the primary credit rate (currently 3.75 percent). Banks borrowing here pay this rate. They rarely do, because it carries stigma.
Standing Repo Facility: a backstop that caps short-rate spikes during stress.

The effective fed funds rate is determined daily by actual transactions and published the next morning by the New York Fed. It is essentially always inside the target range, very close to IORB. That is the rate the Fed controls almost directly.

Everything else, even the 3-month T-bill, is determined by market demand and supply. Yes, the 3-month T-bill is closely tethered to the fed funds rate, but the tether is market arbitrage, not Fed fiat. If the Fed unexpectedly hiked 50 basis points tomorrow, the 3-month T-bill yield would move almost immediately, but it would move because traders sold T-bills and bid up yields, not because the Fed set the T-bill rate.

The Yield Curve Stack

The fundamental decomposition for any Treasury yield with maturity n is:

y_n ≈ (1/n) × Σ E[r_t] + TP_n

In English: the yield on an n-year Treasury equals the average expected short rate over those n years, plus a term premium for that maturity. This is the core insight of the expectations hypothesis, modified by the (well-documented) empirical fact that term premiums are not zero. Two terms, each doing real work:

Average expected short rate: what the market expects the fed funds rate to average over the period. This is sensitive to FOMC communications, inflation data, employment data, and any signal that changes the expected path of policy.
Term premium (TP): extra yield investors require to lock up money for n years rather than rolling short. Compensates for inflation uncertainty, supply-and- demand imbalances in the bond market, and the path-dependent risk of holding a longer-dated security.

Apply this to the Treasury curve as of early May 2026:

2-year Treasury yield: 3.88 percent. This maturity is dominated by expectations. The term premium on a 2-year is small (often 10-30 bps in recent ACM estimates). So 3.88% ≈ ~3.7% expected average short rate + ~0.18% term premium. The 2-year is essentially a market poll on the near-term Fed path.
10-year Treasury yield: 4.39 percent. The NY Fed’s ACM model estimates the 10-year term premium at roughly 0.65 percent in late April 2026 per their published series. That implies the average expected short rate over the next 10 years is roughly 3.74 percent, and term premium adds 0.65 percent.

Notice the 10-year’s expected-short-rate component (3.74%) is lower than the 2-year’s expected-short-rate component (3.7%). They are essentially the same. The difference between the 2-year and the 10-year (3.88% vs 4.39%, a +51 bps spread) is almost entirely term premium, not an expectation that rates will rise. Curve shape carries information that gets lost when commentators talk about “the rate” as if it were one number.

Term Premium: Important and Unobservable

Term premium is critical to understanding the yield curve and also unobservable. We have to estimate it from a model. Different models give different answers. The most cited sources:

Adrian-Crump-Moench (ACM), NY Fed: a five-factor no-arbitrage term-structure model, published daily. Often used in market commentary because it updates frequently.
Kim-Wright, Federal Reserve Board: a three-factor latent variable model, published weekly. Often gives different (sometimes much higher) term-premium readings than ACM.
Crump-Moench-Steele and various academic alternatives: more recent latent-variable approaches.

In late April 2026, ACM’s estimate of the 10-year term premium was around +65 bps. Kim-Wright readings have at times been double that level. Both are model output. Neither is the truth. The point of looking at term premium estimates is not to anchor on a number, but to see when models broadly agree the term premium is high (e.g., late 2023, when supply concerns spiked) versus low (most of 2010-2017, when Fed QE suppressed term premium).

Why this matters for investors. Two scenarios that look identical at the 10-year level (say, 4.4 percent yield) can be very different under the hood:

High expected-rate, low term premium (e.g., 4.0% + 0.4%): short rates are expected to stay elevated. A duration tilt benefits more from carry but is exposed if rates fall.
Low expected-rate, high term premium (e.g., 3.0% + 1.4%): market expects rate cuts but is paying up for the privilege of locking in. Duration benefits more from potential price appreciation if cuts arrive.

Same yield. Different decomposition. Different ex-ante return distribution.

r-star: The Unobservable North Star

Behind both the expected-rate path and the term premium sits r-star, the natural rate of interest. r-star is the real (inflation-adjusted) rate at which monetary policy is neither tight nor loose. Add expected inflation and you get the nominal neutral rate, the level the Fed is trying to find.

r-star is unobservable. The most-cited estimate is the Holston-Laubach-Williams (HLW) model, maintained by the New York Fed, which uses a Kalman filter on data covering output, inflation, and the policy rate to back out a real-time estimate. HLW’s 2026 readings put real r-star around 1 percent in the United States, with uncertainty bands easily +/- 100 bps. With expected long-run inflation near 2 percent, that implies a nominal neutral rate around 3 percent.

Other r-star models (Lubik-Matthes at the Richmond Fed, the Cleveland Fed’s neutral rate dashboard, surveys from Mercatus) report a range from roughly 0.5 to 2 percent real. John Williams, the NY Fed President who has done much of the academic work on r-star, has called it “a star you can’t see”, a useful concept that we estimate badly.

For investors, the takeaway is not the specific number but what r-star uncertainty implies: the Fed itself does not know exactly where neutral is. Confidence intervals are wide. Forecasts of the long-run policy rate (the dot plot’s long-run dot, which is the FOMC’s collective r-star guess plus 2 percent inflation) have drifted between 2.5 and 3.5 percent for years and continue to move. Anyone who tells you the “true” long-run rate is selling something.

Inflation: The Loud Channel

Inflation enters the yield curve in two places:

Expected inflation: nominal yields embed expected inflation. The market’s 5-year breakeven (5-year nominal Treasury yield minus 5-year TIPS yield) was roughly 2.3 percent in early May 2026 per FRED. The 10-year breakeven is similar. These are noisy proxies but the cleanest market-based readings of inflation expectations.
Inflation risk premium: investors demand extra yield for the uncertainty around inflation, not just its expected level. This bleeds into term premium and is often hard to separate cleanly.

When CPI surprises higher than expected, two things tend to happen at once: the market raises its expected path of short rates (the Fed will hold longer or hike), and term premium rises (more uncertainty around inflation worth being paid for). Both push long-end yields up. When CPI surprises lower, both unwind. Most of the day-to-day noise in the 10-year is one of these two channels reacting to data.

Mortgage Rates: Not Just Fed Funds Plus a Spread

The 30-year fixed-rate mortgage averaged 6.30 percent per Freddie Mac’s PMMS as of April 30, 2026. The fed funds rate is 3.50-3.75 percent. The gap is roughly 270 bps. That spread is not a constant. Three components, in order of size:

The 10-year Treasury anchor. 30-year mortgages are priced off intermediate Treasuries (often the 10-year, sometimes the 7-10 blend) because most mortgages prepay or refinance in 7-10 years rather than running their full 30-year stated term.
MBS-Treasury spread (the “basis”). Mortgage-backed securities trade at a yield premium to Treasuries to compensate for prepayment risk: when rates fall, borrowers refinance, and the MBS holder gets cash back earlier than expected at lower reinvestment yields. This spread has ranged from roughly 100 bps in calm periods to over 200 bps during stress (March 2020, late 2022). In early May 2026 the basis is around 130 bps.
Origination, servicing, and credit spread. Lenders charge a margin over the MBS yield to cover origination costs, servicing, and credit risk. Typically 50- 75 bps.

Add them: 4.39 percent (10-year Treasury) + 1.30 percent (MBS-Treasury basis) + 0.61 percent (origination/servicing/ credit) ≈ 6.30 percent. That checks out within rounding.

The implication for borrowers: a Fed rate cut does not translate one-for-one into a mortgage rate cut. If a rate cut is expected, the 10-year may have already adjusted before the cut, and the mortgage rate with it. If the cut surprises the market, the 10-year could even rise (if the cut signals future inflation or supply concerns), pulling mortgage rates higher despite the Fed easing. The Summitward guide Is a 30-Year Fixed Mortgage an Inflation Hedge? treats the borrowing-side question.

Worked Example: Decomposing a Move

Suppose tomorrow the FOMC unexpectedly cuts 25 bps. What happens to the 10-year? The chain runs as follows:

The expected-short-rate component falls. If the cut is read as a one-off, the average expected short rate over 10 years falls by perhaps 5 bps (a 25 bps cut now spread over 120 months matters less than it sounds at the long end). If the cut signals a steeper cutting path, the decline could be 15-20 bps.
Term premium can move either direction. If the cut signals an inflation tolerance, term premium might rise (investors want more compensation for inflation uncertainty), partially offsetting the expected-rate decline. The 10-year might fall less than the cut implies. If the cut signals economic weakness, term premium might fall further (flight to quality), and the 10-year could fall more than the cut implies.
The basis can move with risk appetite. If the cut is positive for risk assets, MBS-Treasury basis can tighten (mortgage rates fall more than the 10-year). If it is negative for risk, the basis widens (mortgage rates fall less).

That is one cut. The chain has at least three moving parts, each pulling its own direction. The simple Fed-cut-equals-mortgage-cut model gets it wrong most of the time.

What This Implies for DIY Investors

Five practical takeaways from the rate-formation chain.

1. Cash yields are not permanent. The 4-week T-bill, money market funds (SGOV, VUSXX), and high-yield savings accounts all track short rates closely. When the Fed cuts, those yields fall within days. Locking in any meaningful chunk of the cash you will need over 1-3 years today, when yields are still elevated, is reasonable. The Summitward guide Where to Park Your Cash walks through which vehicle suits which time horizon.

2. Duration matching beats Fed timing. Liability matching, picking bond duration that matches when you need the money, dominates trying to predict the Fed’s next move. The Summitward guide Asset-Liability Matching covers the framework.

3. Bond fund losses are not permanent impairments. When yields rise, bond fund prices fall. The fund’s yield-to-maturity is the best estimate of future returns over a horizon equal to its duration. If you hold for at least the duration, carry recovers most or all of the price loss. Selling at the bottom is what locks in the loss.

4. Mortgage decisions need yield-curve thinking. Whether to refinance, lock a rate, or buy points depends on where the 10-year is and what the curve is telling you about the future short-rate path. Predicting the Fed’s next 25 bps is rarely the binding question; deciding how confident you are in your hold period and your future mobility is. The Summitward guide Mortgage Points and How Much to Put Down cover the borrowing-side decisions.

5. Equity valuations are sensitive to discount rates. A higher 10-year yield raises the discount rate on future equity cash flows, which compresses growth- stock multiples more than value-stock multiples (longer- duration cash flows are more sensitive). The 2022 selloff in long-duration tech vs the relative resilience of value was the cleanest recent example.

Frequently Asked Questions

Why is the 2-year Treasury called a “market poll” on the Fed?

Because the 2-year is dominated by expected-rate-path math (term premium is small at 2 years), its yield is essentially the market’s collective forecast for where the federal funds rate will average over the next 24 months. When the 2-year yield moves sharply, it is almost always because the market changed its mind about the Fed.

How does QE affect the yield curve?

QE works primarily by compressing term premium. The Fed buys long-dated Treasuries, takes them out of public hands, and forces investors who wanted that duration into other assets, which pushes their yields down too. That is the portfolio-balance channel. QT (quantitative tightening) is the reverse: the Fed lets bonds run off, supply returns to the public market, term premium widens, long yields rise.

Should I buy long bonds because rates will fall?

That is a directional bet on the rate path, and history suggests rate timing is hard. The case for long-duration Treasuries (TLT, EDV) rests more on the asymmetric payoff during deep recessions, when long bonds rally hard, than on rate-call accuracy. They are diversifiers in a stocks-and- bonds portfolio, not a yield play.

Why didn’t mortgage rates fall when the Fed cut in 2024-2025?

Two main reasons. First, the cuts were largely expected, so the 10-year had already moved before the cuts happened, and mortgage rates with it. Second, term premium widened across the cycle as inflation persisted and Treasury supply increased. The expected-rate decline and the term-premium rise partially canceled, leaving the 10-year (and mortgage rates) roughly unchanged even as the front end fell.

Does the Fed coordinate with foreign central banks?

Information-sharing happens through BIS, IMF, and bilateral channels, but each central bank sets its own policy. Diverging cycles (Fed easing while ECB tightens, or vice versa) are common and produce currency moves that affect cross-country bond yields through hedging costs.

Related Guides

Where to Park Your Cash covers SGOV, T-bills, money market funds, and HYSAs and which to use when. Pairs directly with the cash-yield tracker.
Asset-Liability Matching is the framework for choosing bond duration based on when you need the money rather than where you think the Fed is headed.
Is a 30-Year Fixed Mortgage an Inflation Hedge? covers the borrowing-side question through the same yield- curve lens.
Mortgage Points and How Much to Put Down on a House take the rate framework into specific borrowing decisions.
The Best Inflation Hedges Are Boring covers TIPS, I Bonds, and how breakeven inflation embeds in the yield curve.

Key Takeaways

The Fed sets one rate (the FFR target range) and enforces it with two administered rates (IORB at 3.65%, ON RRP at 3.50%). Everything else is set by markets.
y_n ≈ avg(expected short rates) + term premium is the operative decomposition. The 2-year is mostly expectations; the 10-year carries meaningful term premium.
Term premium and r-star are unobservable and model-dependent. Different models give different answers; treat them as estimates, not the truth.
Mortgage rates = 10-year Treasury + MBS-Treasury basis + origination spread. A Fed cut does not translate one-for-one into a mortgage cut.
For DIY investors: cash yields are not permanent (lock duration matched to liabilities), bond fund losses are not permanent impairments (carry recovers them over the duration), and equity valuations are discount-rate-sensitive.