How Much of a Ten-Year Treasury Yield Is Risk Compensation?

Yields move daily. Market levels here are anchored to data available as of August 25, 2026. The framework does not move.

On August 17, 2026 the 30-year Treasury yield closed at 5.31 percent on Treasury’s daily series, its highest since 2007, and it stood at 5.17 percent a week later. The explanation that followed was that bond investors are demanding more compensation to lend to governments for a long time. That compensation has a name, the term premium. It is estimated rather than observed, and the estimate for the United States has barely moved in a year.

The split inside a ten-year yield

A long-term Treasury yield can be decomposed into two pieces: what investors expect short-term rates to average over the life of the bond, and everything else.

Reading the symbols, with the August 24, 2026 values filled in:

• $y_{n}$: today’s yield on a zero-coupon Treasury maturing in $n$ years. For the ten-year, the ACM fitted value was 4.77 percent.
• $n$: maturity in years; 10 for the ten-year.
• $r_{t+i}$: the one-year Treasury rate $i$ years from today, $t$. The sum runs from $i = 0$ (this year’s rate) to $i = n - 1$ (the rate in the bond’s final year).
• $\mathbb{E}_t[\,\cdot\,]$: the market’s expectation today of that future rate.
• $\tfrac{1}{n}\sum \mathbb{E}_t[r_{t+i}]$: the average one-year rate investors expect over the bond’s life; 3.98 percent in the ACM decomposition.
• $TP_{n}$: the term premium for maturity $n$; 0.79 percent.

The $\approx$ is there because the relation holds exactly in continuously compounded yields and ignores a small convexity term; the ACM figures add up exactly in their own units.

The second term is the term premium. The Federal Reserve Bank of New York, which publishes the most widely cited estimate, defines it as “the compensation that investors require for bearing the risk that interest rates may change over the life of the bond.”¹ If you could roll one-year bills for a decade and expect the same return as buying the ten-year outright, the term premium would be zero. The gap between those two strategies is the premium.

Neither piece is quoted anywhere. Both have to be pulled out of observed yields by a model, and the models disagree. The New York Fed uses the Adrian-Crump-Moench specification, a five-factor no-arbitrage model estimated by linear regression on principal components of the yield curve.² Its decomposition of the ten-year on August 24, 2026:

Five sixths of the yield is a forecast of the Fed. One sixth is risk compensation.

Between late August 2025 and late August 2026 the fitted ten-year rose 47 basis points. About 5 of them came from the term premium. The rest, about 42, came from investors raising their forecast of where short-term rates will sit.

Since the end of 2021 the picture is different. The fitted ten-year has tripled, from 1.59 to 4.77 percent, a rise of about 320 basis points that splits roughly 170 from higher expected short rates and 150 from a higher term premium. So the claim that duration risk is repricing holds up over that window and not over the last twelve months. The 30-year yield in the headlines cannot settle the question either way: the New York Fed publishes ACM estimates only out to ten years, so the ten-year is the maturity where the decomposition can be measured at all.

A premium that has been negative half the time since 2010

The intuition that a longer bond has to pay more, because more can go wrong over thirty years than over three, has been false about half the time since 2010. Eric Swanson made the theoretical case at the San Francisco Fed in 2007: there is nothing in the definition requiring a positive number, and investors such as pension funds with long-dated liabilities “could be willing to accept a lower yield on long-term securities (a negative term premium).”³

On the daily ACM file, the ten-year term premium has been negative on 49 percent of trading days since 2010, and on 88 percent of days between 2016 and 2021. It bottomed at negative 1.65 percentage points on March 9, 2020. Its record high was positive 5.25 points on January 11, 1982. The chart plots month-end readings, which smooth those daily extremes to negative 1.36 (July 2020) and positive 5.18 (May 1984).

Today’s 0.79 percent is high by the standards of the years since 2010 and low by the standards of the forty before that: only about one trading day in eleven from 1970 through 2009 carried a ten-year term premium below it. A reader who learned about bonds after 2010 has spent their whole investing life in the most unusual stretch of the series.

One clarification: the New York Fed publishes ACM estimates for maturities of one through ten years only. When commentary refers to “the 30-year term premium,” it is either quoting the ten-year figure or using a different model.

What the premium is compensation for

Someone buying a 30-year nominal Treasury is committing today’s dollars to a contract whose value depends on three decades of unknown inflation, policy, growth, bond supply, and the willingness of other people to hold the same paper. Several forces push on the price of that commitment, and the research behind each varies a lot in strength.

The inflation row deserves a caveat that its own literature supplies. Wright’s results drew a published Comment in the same journal from Michael Bauer, Glenn Rudebusch, and Cynthia Wu, arguing that term premia from maximum-likelihood estimates of affine models suffer small-sample bias severe enough to render his estimates “essentially acyclical.” Wright replied that the bias-corrected figures sit worse against survey evidence.⁸ The argument is unresolved. Related small-sample concerns apply to affine term structure models generally, though ACM’s regression-based estimator is not the maximum-likelihood setup that Comment examined.

Expected inflation and inflation uncertainty are also separate channels, and only the second one belongs in the term premium. If everyone became convinced inflation would run at 3 percent forever and were completely certain of it, nominal yields would rise because expected short rates rise. The premium for bearing risk need not move at all. On August 25, 2026 the ten-year breakeven sat at 2.32 percent against a ten-year TIPS real yield of 2.32 percent, so on that day the nominal long rate was carried by the real yield rather than by expected inflation. Two cautions: a breakeven also contains an inflation risk premium and a TIPS liquidity discount, and that 2.32 percent expectation sits well below the 3.5 percent inflation the S&P Cotality release quoted for June.

Deficits, supply, and who absorbs the duration

Government spending does not mechanically raise the term premium. What the bond market has to absorb is the financing: how much duration ends up in private hands, at what maturity, and who is willing to hold it.

Thomas Laubach’s Federal Reserve study, which used projected rather than contemporaneous deficits to blunt the reverse-causality problem, put the effect at roughly 25 basis points on the ten-year rate expected five years ahead per percentage point of projected deficit-to-GDP, with another 3 to 4 basis points per point of projected debt-to-GDP.⁹ That is an estimate for the whole long rate, not for the premium component alone.

Treasuries also pull the other way. Krishnamurthy and Vissing-Jorgensen’s 73 basis point convenience yield exists because Treasuries are useful as collateral, as regulatory capital, and as the asset everyone runs to. Issuing more of them satisfies some of that demand and erodes the discount. So more debt raises the compensation needed to absorb duration and simultaneously reduces the scarcity that was suppressing yields. Which effect dominates has to be estimated; it cannot be reasoned out.

US supply and ownership as of August 2026, from Treasury’s and the Fed’s own books:

The composition change is the piece with a direct theoretical link to the premium. At its June 2026 meeting, the Federal Open Market Committee heard from the desk manager that “the ownership composition of Treasury securities has shifted somewhat over the past several years from relatively price-insensitive official-sector holders to more price-sensitive private investors, which could have implications for the term premium component of yields.”¹² That was a market-desk observation rather than a Committee judgment, but the Treasury’s own capital-flows data backs it: over the year to June 2026, foreign private investors added $319 billion while foreign official institutions sold $114 billion.

Reserve managers buy Treasuries for policy reasons and hold through price moves. Asset managers and leveraged funds demand a yield to hold the same paper and step away when it is not paid. Shifting the marginal buyer toward the second group makes the market more price-sensitive, which is a coherent reason to expect a higher and more variable premium than the 2010s delivered.

Three current facts cut against the loudest version of the fiscal story. Bills sit at 22.2 percent of marketable debt, which is above the 15 to 20 percent range the Treasury Borrowing Advisory Committee recommended in 2020 and the roughly 20 percent its 2024 review settled on, and at the same time right at the 22.4 percent average since 1980; “flooding the front end” is a stretch. The Fed has stopped shrinking its Treasury book, adding about $27 billion over the past month, though every dollar went into bills rather than duration. And on August 19, 2026, Treasury doubled its long-end buyback operations from $2 billion to at least $4 billion apiece in the 10-to-20 and 20-to-30 year sectors, effective September 9.¹³

What it costs a household

In mid-September 2024 the federal funds target was 5.25 to 5.50 percent. In August 2026 it is 3.50 to 3.75 percent, 175 basis points lower. The week the Fed made its first cut, Freddie Mac’s 30-year fixed mortgage rate was 6.09 percent. In the week ending August 20, 2026 it was 6.65 percent. Nearly two percentage points of easing, and the mortgage rate rose 56 basis points.

Dallas Fed researchers quantified why in May 2026. Regressing the 30-year current-coupon mortgage spread over the ten-year Treasury on three factors, the level of ten-year rates, the fed funds to ten-year curve slope, and swaption implied volatility, they explained about 70 percent of the spread’s variation over twenty years. Their headline result:

“The mortgage rate exhibits a partial beta of less than 20 percent with respect to the fed funds rate, while exhibiting an 85 percent beta with respect to the 10-year rate.”¹⁴

Partial betas describe co-movement in a regression rather than a transmission mechanism, but the gap is large. And the ten-year does not follow the Fed closely. The same paper reports that over twenty years it moved about 32 percent as much as fed funds, over the last five years closer to 15 percent, and that the relationship “has even turned negative the past one to two years.”¹⁴

The paper also sets the right benchmark. A 30-year mortgage amortizes, so it behaves like a weighted ladder of loans from one to thirty years rather than a single 30-year bond. At mortgage rates between 3 and 7 percent, the Treasury with the same duration is a seven to ten year Treasury, which is why the industry quotes mortgage spreads against the ten-year.

Run the arithmetic yourself

This calculator applies those two coefficients to today’s market. Move the Fed with the first slider and move the long end on its own with the second. The second slider is where the term premium shows up, and the Fed has no direct lever on it.

Four rows from the tool. All of them assume a $400,000 loan and the August 25, 2026 starting point of a 4.64 percent ten-year and a 6.65 percent mortgage, and they would change with different inputs:

For this borrower, a two-point Fed cutting cycle is worth about $16 a month if long rates rise a point against them, while the Fed doing nothing at all and the long end falling a point is worth about $222 a month. The difference between those two outcomes is entirely the long end.

The buying-power version of the same arithmetic, on the same $400,000 loan: a one-point drop in the ten-year reaches the mortgage as about 86 basis points and buys about $38,000 more house at an unchanged payment. A full point of mortgage-rate relief buys about $45,000 more; a full point against the borrower takes about $38,000 away. The relationship is convex, so relief helps more than the same move hurts.

What higher rates did to prices, sales, and rents

San Francisco Fed researchers measured the listing-price response around monetary policy announcements. They found that “a 1 percentage point increase in the mortgage rate generated by a monetary policy shock is associated with house prices falling immediately by 1%,” settling “3% lower three weeks after the shock.”¹⁵ They also found the channel that matters here: “House prices respond more strongly to unexpected changes in long-term interest rates than to surprises in the short-term federal funds rate.”

Prices, though, absorbed less of the shock than transaction volume did, because owners with cheap mortgages stopped selling. FHFA staff estimated that every percentage point by which market rates exceed a borrower’s origination rate cuts their probability of selling by 18.1 percent, which produced a 57 percent reduction in fixed-rate home sales in the fourth quarter of 2023 and prevented 1.33 million sales between mid-2022 and the end of 2023. The lost supply pushed prices up 5.7 percent, more than offsetting the 3.3 percent decline that higher rates caused directly.¹⁶

Lock-in is still with us. As of the first quarter of 2026, 78 percent of outstanding US mortgages carried a rate below 6 percent, and 19.5 percent were still below 3 percent; the average outstanding rate was 4.4 percent. Against the August 2026 market rate of 6.65 percent, the typical borrower sits about 225 basis points under the market.

Lock-in reaches renters through the same channel. Modeling it in general equilibrium, Fonseca, Liu, and Mabille estimate that a temporary rate increase in their calibrated model raises aggregate house prices 4.4 percent and rents 1.5 percent relative to a world without lock-in, while cutting borrower mobility 25 percent. Households who would have downsized stay in larger homes instead, which keeps housing demand high in exactly the places it is most expensive.¹⁷ In their separate credit-record study, Fonseca and Liu found each percentage point of rate disadvantage reduces the probability of moving by 0.68 percentage points, or about 9 percent. That is a smaller elasticity than FHFA’s 18.1 percent because it counts all moves, including moves that keep the mortgage, rather than sales alone.

The cumulative damage shows up in affordability. In 2020 the National Association of Realtors put the mortgage payment on a median-priced home at 14.7 percent of median family income, or $1,035 a month. In July 2026 it was 24.2 percent, or $2,254. The payment more than doubled while the median price rose 47 percent; price and rate each contributed a factor of roughly 1.5. The Atlanta Fed’s broader measure, which includes taxes, insurance, and mortgage insurance at 10 percent down, had the median household spending 43.4 percent of income to own the median home in May 2026, against a conventional 30 percent threshold, 29.1 percent in mid-2019, and a record 45.2 percent in October 2023.²¹

And prices themselves never gave much back. The S&P Cotality Case-Shiller national index fell 5.0 percent from its June 2022 peak into January 2023 and then recovered all of it: the June 2026 reading of 336.66 sits 9.3 percent above that 2022 peak. Nominal prices are still rising, at 1.5 percent year over year, and with the 3.5 percent inflation quoted in the same release, real house prices are falling about 2 percent a year.

What the predictability evidence survived

Time-varying bond risk premia are real and well documented. Campbell and Shiller showed in 1991 that the expectations hypothesis fails in postwar US data in a specific way: when the yield spread is high, the long rate tends to fall, which is the opposite of what the theory predicts and is consistent with premia that move around.¹⁸ Cochrane and Piazzesi later found that a single tent-shaped combination of forward rates predicts one-year excess returns on one- to five-year bonds with an R-squared “up to 0.44.”¹⁹

Then the statistics got a second look. Michael Bauer and James Hamilton showed that the tests underlying much of this literature suffer serious small-sample distortions, built a bootstrap designed for the question, and revisited six published studies. Their finding was that the evidence against the spanning hypothesis, the idea that the current level, slope, and curvature of the yield curve already contain everything predictive, is “much weaker than it originally appeared.”²⁰ Two of the six are Cochrane and Piazzesi (2005), cited in the paragraph above, and Greenwood and Vayanos (2014), the supply evidence cited earlier. Both results carry that caveat.

The measurement problem compounds it. In June 2007, four reasonable methodologies produced ten-year term premium estimates ranging from negative 2 percent to positive 2 percent, which Swanson called “a tremendous difference considering that the 10-year Treasury yield has been only about 5%.”³ Model disagreement of that size rules out any trading rule keyed to a single estimate crossing a threshold. Term premia exist and expected bond returns vary over time; pinning down today’s number, or forecasting tomorrow’s, is a long way behind both.

What to do with this

A higher term premium improves the prospective compensation for owning duration. It does nothing to the risk. A long-Treasury fund with a 16-year duration still loses roughly 16 percent of its price on a one-point rise in yields, a little less after convexity. Ask whether you have a reason to own this much duration before asking whether the premium looks attractive.

1. Match duration to the purpose of the money. Money you will spend within three years belongs in cash and bills. Beyond that, hold bonds whose duration is roughly the number of years until you need the money, and do not extend it because a 30-year yield looks tempting.
2. Use intermediate high-quality bonds as the default ballast. A total-bond-market or intermediate Treasury fund runs a duration of about five to seven years. That is enough to matter in a recession without turning the portfolio into a position on the long end.
3. Reach for long Treasuries deliberately. Fifteen-plus years of duration earns a place when you are hedging a specific long-dated liability, such as a pension you are funding yourself, or when you want the strongest available recession and deflation hedge and can sit through drawdowns of the kind long Treasuries took from 2020 to 2023.
4. Use TIPS when the liability is real purchasing power. Retirement spending is a real liability. Nominal long bonds hedge inflation indirectly and badly; a TIPS ladder or fund matched to the same horizon hedges it directly.
5. Do not time duration off a single term premium model. ACM and Kim-Wright routinely differ by more than the move you would be trading on.
6. Stop treating a Fed cut as a mortgage forecast. If you are waiting to buy a house until the Fed eases, you are watching the wrong rate. Decide on your hold period, your budget, and your job stability; those are answerable.

Long rates are also discount rates, so the same decomposition reaches equity valuations. The rate-formation guide works through that channel and the rest of the yield curve machinery.

Key takeaways

• Most of a long yield is a Fed forecast. On August 24, 2026 the New York Fed’s model split a 4.77 percent ten-year into 3.98 percent of expected short rates and 0.79 percent of term premium. Over the past year the fitted yield rose 47 basis points and the term premium about 5 of them; since the end of 2021 the split is closer to even.
• The premium can be negative, and often was. The ACM ten-year term premium was below zero on 49 percent of trading days since 2010 and on 88 percent of days from 2016 to 2021.
• Deficits reach long rates through financing. What moves them is how much duration private investors must absorb, at what maturity, and whether the buyer cares about price. Foreign official holders have been shrinking while private ones grow.
• Your mortgage tracks the long end. Dallas Fed estimates put the mortgage rate’s partial beta at 85 percent to the ten-year Treasury and under 20 percent to fed funds. Since the Fed’s first cut in September 2024, fed funds is down 175 basis points and the 30-year fixed is up 56.
• Lock-in reaches renters too. A calibrated general-equilibrium model puts the effect of a temporary rate increase at 4.4 percent on house prices and 1.5 percent on rents, with borrower mobility down 25 percent.
• Do not trade on the estimate. Four methods spanned negative 2 to positive 2 percent on the same day in 2007,³ and the robustness literature has weakened much of the predictability evidence.

How Summitward helps

Frequently asked questions

Is the yield curve slope the same thing as the term premium?

No. The gap between the ten-year and the three-month yield contains both the market’s expectation that future short rates will differ from today’s and the term premium. Subtracting two observed yields cannot separate them, which is why the New York Fed publishes a model-based estimate rather than a subtraction.

How do the ACM and Kim-Wright estimates differ?

ACM extracts five principal components from the yield curve and estimates the model by a three-step linear regression, published daily. Kim-Wright, maintained at the Federal Reserve Board, uses three latent factors estimated by maximum likelihood and is published weekly. The estimation method is where the small-sample critique bites hardest, and it is one reason the two can disagree by half a percentage point on the same day. Treat either as one model’s reading.

Does a rising term premium mean the bond market is worried about default?

Not by itself. A higher long yield can reflect a higher expected path of short rates, more inflation uncertainty, more duration supply, less Treasury scarcity, changing foreign demand, constrained dealer balance sheets, or a shift in how well bonds hedge stocks. Sovereign credit concern is one candidate among many, and the current ACM estimate of 0.79 percent is low against the series’ sixty-five-year history.

What would make the term premium fall from here?

The same forces in reverse: lower inflation uncertainty, lower interest rate volatility, a return of price-insensitive buyers such as foreign reserve managers or the Fed buying coupons rather than bills, or bonds again reliably rallying when stocks fall. The calculator’s second slider shows what a given amount of that compression would do to a mortgage payment without the Fed moving at all.

Was the negative term premium a free lunch for borrowers?

For anyone who locked a 30-year fixed rate then, yes. Freddie Mac’s survey rate bottomed at 2.65 percent in January 2021, a level that embedded a ten-year term premium of about negative one point. Those borrowers were paid to take a rate the market expected to lose money on, and the lock-in that now freezes the housing market is the mirror image of that bargain.

What would bring mortgage rates down?

A lower ten-year Treasury yield, a narrower mortgage-backed securities spread, or both. The spread narrows when rate volatility falls or when price-insensitive buyers step in. The ten-year falls when expected short rates fall or when the term premium compresses. Fed cuts influence the first of those and can coincide with the opposite of the second.

Related guides

• Does the Fed Really Set Interest Rates?: the full rate-formation chain, r-star, and what the Fed does and does not control.
• Are Bonds Still Good Diversifiers?: why the stock-bond correlation flips sign, and what that does to the case for duration.
• Austria’s 100-Year Bond: what duration and convexity do to a price at extreme maturity.
• HEDGEFUNDIE’s Excellent Adventure: how much a levered long-Treasury strategy depends on the term premium assumption you feed it.
• Is a 30-Year Fixed Mortgage an Inflation Hedge?: the same yield curve seen from the borrower’s side of the contract.
• The Risk-Free Rate Is Your Hurdle Rate: how to use Treasury and TIPS yields, and what a breakeven does and does not tell you.

Sources

1. Federal Reserve Bank of New York, Treasury Term Premia. Daily and monthly ACM estimates for maturities of one to ten years, 1961 to present. Figures in the text are from the daily file through August 24, 2026; the chart plots the month-end file.
2. Tobias Adrian, Richard K. Crump, and Emanuel Moench, “Pricing the term structure with linear regressions”, Journal of Financial Economics 110(1), 2013, pp. 110-138.
3. Eric Swanson, “What We Do and Don’t Know about the Term Premium”, FRBSF Economic Letter 2007-21, July 20, 2007.
4. Jonathan H. Wright, “Term Premia and Inflation Uncertainty: Empirical Evidence from an International Panel Dataset”, American Economic Review 101(4), 2011, pp. 1514-1534.
5. Richard H. Clarida, “Monetary Policy, Price Stability, and Equilibrium Bond Yields: Success and Consequences”, speech, November 12, 2019. The 100 basis point figure is attributed in the speech to the yield curve model of Don Kim, Cait Walsh, and Min Wei.
6. Robin Greenwood and Dimitri Vayanos, “Bond Supply and Excess Bond Returns”, Review of Financial Studies 27(3), 2014, pp. 663-713.
7. Brian Bonis, Jane Ihrig, and Min Wei, “The Effect of the Federal Reserve’s Securities Holdings on Longer-term Interest Rates”, FEDS Notes, April 20, 2017.
8. Michael D. Bauer, Glenn D. Rudebusch, and Jing Cynthia Wu, “Term Premia and Inflation Uncertainty: Comment”, American Economic Review 104(1), 2014, pp. 323-337, with Wright’s Reply at pp. 338-341.
9. Thomas Laubach, “New Evidence on the Interest Rate Effects of Budget Deficits and Debt”, FEDS Working Paper 2003-12; published in the Journal of the European Economic Association 7(4), 2009, pp. 858-885. The debt-to-GDP coefficient is 3 to 4 basis points in the published version and 4 to 5 in the working paper.
10. Arvind Krishnamurthy and Annette Vissing-Jorgensen, “The Aggregate Demand for Treasury Debt”, Journal of Political Economy 120(2), 2012, pp. 233-267.
11. Srini Ramaswamy, Seth Searls, Hugo De Vere, and Ipek Ozil, “Term funding premium: Time is money even absent interest rate risk”, Federal Reserve Bank of Dallas, June 25, 2026. Their decomposition is term premium = term funding premium + term rate premium; the regressions in their Table 1 cover January 2019 through December 2025.
12. Federal Open Market Committee, Minutes of the meeting of June 16-17, 2026. The quoted passage appears in the desk manager’s report under Developments in Financial Markets and Open Market Operations.
13. U.S. Department of the Treasury, “Treasury Announces Increased Sizes of Nominal Long-End Liquidity Support Buybacks Beginning September 9”, August 19, 2026. Holdings and supply figures draw on Treasury Debt to the Penny, Treasury International Capital Table 5, Federal Reserve statistical release H.4.1, and Treasury’s Q3 2026 presentation to the Treasury Borrowing Advisory Committee. The 2008 foreign-share peak is computed from the Federal Reserve’s Financial Accounts series for foreign holdings against debt held by the public, second quarter 2008. TBAC’s bill-share guidance is from its November 2020 recommendation and its third-quarter 2024 charge on T-bill issuance.
14. Matthew McCormick and Srini Ramaswamy, “What drives mortgage rates and their response to monetary policy changes”, Federal Reserve Bank of Dallas, May 7, 2026.
15. Denis Gorea, Augustus Kmetz, Oleksiy Kryvtsov, Marianna Kudlyak, and Mitchell Ochse, “House Prices Respond Promptly to Monetary Policy Surprises”, FRBSF Economic Letter 2023-09, March 27, 2023. The underlying data is listing prices.
16. Ross M. Batzer, Jonah R. Coste, William M. Doerner, and Michael J. Seiler, “The Lock-In Effect of Rising Mortgage Rates”, FHFA Staff Working Paper 24-03, March 18, 2024. Outstanding-mortgage rate distributions are from the FHFA National Mortgage Database, first quarter 2026.
17. Julia Fonseca, Lu Liu, and Pierre Mabille, “Unlocking Mortgage Lock-In: Equilibrium Effects in a Spatial Housing Ladder Model”, NBER Working Paper 35237, 2026; and Julia Fonseca and Lu Liu, “Mortgage Lock-In, Mobility, and Labor Reallocation”, Journal of Finance 79(6), 2024.
18. John Y. Campbell and Robert J. Shiller, “Yield Spreads and Interest Rate Movements: A Bird’s Eye View”, Review of Economic Studies 58(3), 1991, pp. 495-514.
19. John H. Cochrane and Monika Piazzesi, “Bond Risk Premia”, American Economic Review 95(1), 2005, pp. 138-160.
20. Michael D. Bauer and James D. Hamilton, “Robust Bond Risk Premia”, Review of Financial Studies 31(2), 2018, pp. 399-448.
21. Affordability, price, and rate data: National Association of Realtors Housing Affordability Index (July 2026, preliminary), the Atlanta Fed Home Ownership Affordability Monitor (May 2026), S&P Cotality Case-Shiller indices (June 2026, released August 25, 2026), and Freddie Mac’s Primary Mortgage Market Survey (6.09 percent for the week of September 19, 2024; 6.65 percent for the week ending August 20, 2026; the 2.65 percent low was January 7, 2021). Treasury yields, including the 30-year’s 5.31 percent close on August 17, 2026 and its comparison against every year back to 2007, are from Treasury’s daily yield curve.

Author disclosure

Educational content, not investment, tax, or legal advice. Term premium figures are model estimates published by the Federal Reserve Bank of New York and are not observed market prices. The mortgage scenarios use published Dallas Fed coefficients applied to market levels on a single date and are sensitivity exercises rather than forecasts. Your own results depend on your loan size, credit, timeline, and the rates available to you.