Are TIPS Safer Than They Seem? We Re-Ran Elm's Spending-Risk Test

Robert Merton spent part of his 2023 Rational Reminder interview on a point about retirement accounts that sounds obvious until you look at how they are reported. “The risk is your income in retirement, not the value of the account.” For a saver with a long horizon, he said, the risk-free asset “is going to be something like [TIPS]... adjusted to the duration of when you’re planning to retire, not a treasury bill, which is probably one of the riskiest assets you could hold in a retirement plan.” (The retirement segment starts at 56:47.)

Most investors measure the other way. A 30-year Treasury Inflation-Protected Security fell about 42% after inflation in the twelve months to September 2022, in our model, while a T-bill barely moved. On a brokerage statement, the bill was the safe one.

On October 14, 2026, Victor Haghani and James White of Elm Wealth published “TIPS: Safer Than They Seem”, which puts a number on Merton’s argument. They measured wealth as the level inflation-adjusted spending it could buy over 50 years, and reported that by that measure “TIPS were by far the least risky. Nominal bonds were more than twice as risky, and T-bills, which have nearly no market-value volatility, were five times as risky.” We rebuilt their test from public data to see whether it holds up, where it depends on choices the authors made, and what it means for someone deciding how to fund their own retirement spending.

Two Ways to Measure the Same Bond

The usual yardstick is market value. If a holding is worth $W_t$ after inflation, its risk is how much $W_t$ moves from year to year.

Elm’s yardstick asks what that wealth buys. Let $A_t$ be the price today of one dollar of inflation-adjusted spending every year for $N$ years, discounted at the real yield $r_t$ :

A_t = \sum_{k=1}^{N} \frac{1}{(1+r_t)^{k}} = \frac{1-(1+r_t)^{-N}}{r_t}, \qquad c_t = \frac{W_t}{A_t}

Here $c_t$ is the level real spending the holding could fund. Over one year its change is

\frac{c_{t+1}}{c_t} - 1 = (1+R_{t+1})\,\frac{A_t}{A_{t+1}} - 1

where $R_{t+1}$ is the holding’s real return. When real yields fall, $A_t$ rises and each dollar buys less future spending. A long TIPS gains in price at the same moment, so the two terms offset. A T-bill has no price gain to offer, so the full rise in $A_t$ lands on the bill holder’s spending. That is the whole mechanism behind Merton’s claim that bills are risky for retirement savers.

The Replication: TIPS Varied Least in Spending Terms

Elm compared three Treasury holdings over years ending September 30, starting with 1999: T-bills, 30-year nominal bonds and 30-year TIPS, rolling into a new 30-year bond each year. We did the same with the Federal Reserve’s fitted TIPS and nominal yield curves, the Treasury’s daily par real yield curve, 3-month bill rates and CPI-U. The Fed’s TIPS curve begins in January 1999, so our first year ends in September 2000, one year after Elm’s. Toggle the chart between the two yardsticks.

Holding	Spending change, mean (Elm)	Spending change, std. dev. (Elm)	Real market value, std. dev.
30-year TIPS	1.3% (1.3%)	2.3% (2.8%)	14.7%
30-year Treasury	1.0% (0.7%)	6.3% (6.7%)	12.6%
T-bills	−0.1% (0.2%)	14.7% (14.2%)	2.3%

Summitward replication, years ending September 30, 2000 to 2026 (27 years). Elm’s figures, in parentheses, are from the table printed on their chart and cover 1999 to 2026. No taxes or trading costs. Script and data: summitward-research.

The ranking reproduces, and so do the magnitudes. Measured in market value, the 30-year TIPS was the most volatile of the three and the T-bill the least, with standard deviations of 14.7% and 2.3% a year. Measured in spending, the order flips with almost the same numbers: 2.3% for TIPS and 14.7% for bills. Over the full period the spending a rolled TIPS position could fund rose 40%, nominal bonds 26%, and T-bills fell 25%.

The result also survives a cleaner subsample. Using only years where Treasury’s own 30-year real yield exists at both ends (2011 to 2026), the spending standard deviations were 2.6% for TIPS, 6.8% for nominal bonds and 18.0% for bills.

Where our version differs from Elm’s

One fewer year. We start a year later because the Fed’s TIPS curve begins in 1999. That likely explains most of the gap in average T-bill results.
The long real yield before 2010. Treasury did not issue 30-year TIPS between 2001 and 2010 and publishes no 30-year real yield before February 2010. Elm used “the longest maturity bond available.” We used the Fed’s fitted 20-year TIPS yield, the longest it publishes, and priced a 30-year bond off it.
Flat-yield pricing. Each bond is priced at a single yield, and the 50-year annuity is discounted at the same long real yield. A full yield curve would change the individual years.
The CPI lag is in. TIPS principal follows the reference CPI, which under Treasury’s rules is the CPI “for the third preceding calendar month,” interpolated daily. We accrue principal on that index and deflate by the actual September-to-September CPI, so the lag shows up in the results.

A 30-Year Bond Hedges About 50 Years of Spending

Elm paired the 30-year bond with a 50-year spending horizon, writing, “Although a 30-year bond doesn’t perfectly match our assumed 50-year spending horizon, it still provides a useful comparison.” The pairing turns out to be a close one. To see how much it contributes, we reran the same holdings against a 30-year spending stream, the horizon a 65-year-old is more likely to plan for. The TIPS spending standard deviation rose from 2.3% to 6.7%, about the same as nominal bonds (6.6%), and its worst year went from −8.1% to −22.7%.

The reason is duration, the percentage price change for a one-point move in yield. A bond pays most of its value back at maturity, while a spending stream pays evenly, so a 30-year bond carries far more rate sensitivity than 30 years of spending:

Real yield	30-year TIPS at par	50 years of spending	30 years of spending
0%	29.5	25.5	15.5
1%	25.9	23.3	14.7
2%	22.5	21.2	13.9
3%	19.7	19.2	13.1

Modified duration in years at a flat real yield. The 0% row uses Treasury’s 0.125% minimum TIPS coupon. Source: Summitward calculation.

At a 2% or 3% real yield, a single 30-year TIPS has almost exactly the duration of 50 years of level spending, so the pairing in Elm’s test is close to a true hedge. At a zero yield, the bond’s duration (29.5) overshoots the 50-year stream’s (25.5). That gap is Elm’s own explanation for 2022, when real yields rose from about zero and “the bond lost more than the annuity it was hedging.” In our version, 2022 was the worst TIPS year by a wide margin (−8.1%, against −2.0% for the next worst, 2013); excluding it, the TIPS standard deviation drops to 1.4%.

For a household, the lesson is about construction. Thirty years of spending is matched by thirty bonds, one maturing each year, which is what a TIPS ladder is. A single long bond or a long-duration TIPS fund hedges a longer, later stream than most retirees have, even when its maturity equals the date of the last payment.

The Same Idea for One Future Bill

The annuity math generalizes to any dated expense. Suppose you owe $100,000 of today’s purchasing power in 20 years. On October 14, 2026 the 20-year TIPS yield on Summitward’s markets page was 3.09%, so the bill’s present value was

L_0 = \frac{\$100{,}000}{(1.0309)^{20}} = \$54{,}409

If real yields fall one point, that same bill is worth $66,120, 21.5% more. A T-bill holder who set aside $54,409 now needs about $11,700 more to stay funded. A zero-coupon TIPS maturing in year 20, bought for $54,409, rises by the same 21.5%, and its owner is still exactly funded. Pension accountants call the difference between assets and the present value of what they must pay the surplus, $S = A - L$ , and a second-order expansion shows what keeps it steady when yields move by $\Delta y$ :

\Delta S \approx -\left(D_A A - D_L L\right)\Delta y + \tfrac{1}{2}\left(C_A A - C_L L\right)(\Delta y)^2

$D$ is modified duration and $C$ is convexity, for assets $A$ and liabilities $L$ . Two things follow. First, durations must be weighted by dollars: equal durations protect the surplus only when assets and liabilities are also equal in value. Second, the duration term only cancels small moves. Buying a bond whose cash flows land on the dates and in the amounts of the bills cancels every term, which is why cash-flow matching is sturdier than duration matching.

Taxes Weaken the Hedge in a Taxable Account

Every number above is before tax, and TIPS have an awkward tax treatment. Under IRS Publication 1212, a holder “must report as OID any increase in the inflation-adjusted principal amount” each year, while the cash for that increase arrives only at maturity or sale. Treasury interest is exempt from state and local income tax, which helps in high-tax states.

With real yield $r$ , inflation $\pi$ and a marginal rate $\tau$ applied to the whole nominal return each year, the after-tax real return is

r_{\text{after tax}} = \frac{\big[(1+r)(1+\pi)-1\big](1-\tau) - \pi}{1+\pi}

Take a 3% real yield, 4% inflation and a 24% federal bracket. The nominal return is 7.12%, and after tax and inflation the holder keeps 1.36% a year. The tax on inflation compensation takes more than half of the real yield.

Inflation	TIPS, 24% bracket	Nominal, 24% bracket	TIPS, 35% bracket	Nominal, 35% bracket
0%	2.28%	4.24%	1.95%	3.62%
2.5%	1.69%	1.69%	1.10%	1.10%
4%	1.36%	0.23%	0.60%	−0.36%
6%	0.92%	−1.66%	−0.03%	−2.24%
8%	0.50%	−3.48%	−0.64%	−4.05%

After-tax real return, hypothetical. TIPS real yield 3%. The nominal bond yields 5.575%, which equals the TIPS return when inflation is 2.5%. Tax applied to the full nominal return each year; ignores state tax, timing and basis effects. Source: Summitward calculation.

The table makes two points. Nominal Treasuries are taxed on their inflation compensation too, so at the breakeven inflation rate the two bonds leave the same after-tax return. The difference is exposure: each extra point of inflation costs a taxable TIPS holder roughly $\tau$ points of real return (about a quarter point in the 24% bracket), against about one full point for a nominal bond holder. A taxable TIPS still hedges most of an inflation surprise, but it no longer locks in a real return. At a 24% rate, the after-tax real return on a 3% TIPS reaches zero once inflation passes about 10.5%; at 35%, about 5.9%.

The tax can also exceed the cash. A TIPS with Treasury’s minimum 0.125% coupon in a year of 8% inflation pays about 0.14% of par in interest and generates a 24%-bracket tax bill of about 1.95% of par. Many TIPS issued in 2020 and 2021 carried that coupon. Elm’s advice is the standard one: hold TIPS in a tax-deferred account where possible. The TIPS ladder guide covers the account-location order.

The Objections, Taken One at a Time

Most of the common arguments against TIPS are correct for a narrower case than the one they are usually applied to.

“TIPS lose money when rates rise.” Correct, for anyone who sells. PIMCO’s 15+ year TIPS fund returned −31.60% in 2022. It does not change the inflation-adjusted cash a held bond pays, and a household that may need to sell before maturity is exposed to exactly this risk.
“TIPS track CPI-U, not my inflation.” Correct. Principal is indexed to the non-seasonally adjusted CPI for All Urban Consumers. A retiree whose spending tilts toward health care or housing can see prices rise faster than the index. TIPS hedge broad inflation; the gap between your basket and CPI-U stays with you.
“The adjustment lags inflation.” Correct, by about three months. Over a 20-year holding the lag shifts the index window rather than skipping inflation, and our replication includes it. It matters most to someone buying or selling around a sharp, short spike.
“TIPS are tax-inefficient.” Correct in a taxable account, as the tax table shows, with the caveat that nominal bonds pay the same tax at breakeven inflation. The real difference is that TIPS tax rises with realized inflation.
“Stocks or gold are better long-term inflation hedges.” Stocks have a higher expected return, which is a reason to own them for spending you can flex. Neither stocks nor gold promise a specific amount of purchasing power on a specific date, which is the job a matched TIPS does. See why the best inflation hedges are boring.
“The government could fudge the CPI.” Possible, and the contract only partly protects you. Treasury’s rules require it to replace the CPI with an alternative index if the CPI is discontinued or, “in the judgment of the Secretary, fundamentally altered in a manner materially adverse to the interests of an investor,” or altered that way by legislation or executive order. The same rules say “Determinations of the Secretary in this regard will be final.” Elm judges the risk of a politically forced rewrite to be low. Owning TIPS is a bet on U.S. institutions either way, as is owning a nominal Treasury.
“A 30-year ladder runs out.” Correct. Merton noted in the same interview that “there are not real annuities I’m aware of that can be purchased.” Social Security is the main inflation-indexed life annuity most households have, and delaying it buys more of it.

What the TIPS Market Tells Investors Who Never Buy It

Merton opened the interview with market efficiency: “the market has information, very important information that no one has.” A traded real yield curve is a case in point. Before TIPS, an investor’s real discount rate was a model output. Now the price of a future dollar of purchasing power is quoted daily, which is what lets Summitward price a retirement income floor at market.

The quote is not a clean reading of the real risk-free rate. Federal Reserve economists Kim, Walsh and Wei decompose it as

y^{\text{TIPS}} = r^{\text{real}} + LP^{\text{TIPS}}, \qquad y^{\text{nom}} - y^{\text{TIPS}} = \mathbb{E}[\pi] + IRP - LP^{\text{TIPS}}

where $LP^{\text{TIPS}}$ is a liquidity premium for TIPS trading less easily than nominal Treasuries and $IRP$ is an inflation risk premium. D’Amico, Kim and Wei estimate that TIPS yields “exceeded risk-free real yields by as much as 100 basis points when TIPS were first issued and up to 300 basis points during the 2007–2008 financial crisis,” mostly for that reason. Those figures are estimates from a term-structure model. Two practical consequences follow. A liquidity premium is a cost to sellers and extra yield for a buyer who holds to maturity. And the gap between nominal and TIPS yields, 2.35 points at ten years on October 14, 2026, mixes expected inflation with premiums; the Fed’s own TIPS curve file warns that breakeven rates “should not be interpreted as estimates of inflation expectations.”

For what a real yield implies about stock returns, see how real yields set the equity hurdle and the risk-free hurdle rate.

When TIPS Are the Wrong Tool

Spending in the next year or two. Cash and T-bills match a near, nominal bill. Their reinvestment risk only matters over long horizons.
Fixed-dollar obligations. A mortgage payoff or a contract priced in dollars is a nominal liability, and a nominal Treasury maturing on the date matches it better.
Money you may have to sell early. A long TIPS sold after real yields rise realizes the price loss the spending yardstick ignores.
Households that cannot fund a floor. If essential spending costs much more than the portfolio, locking in a fraction of it can lower the odds of meeting the rest. Our income floor analysis puts that crossover near a 0.65 funded ratio in Summitward’s simulation model.
Spending in another currency. TIPS pay dollars indexed to U.S. prices, so a retiree spending euros or pesos carries currency risk on top.

Key Takeaways

The yardstick decides which Treasury looks risky. In our replication of Elm Wealth’s test (years ending September 30, 2000 to 2026), rolled 30-year TIPS had a 14.7% standard deviation in real market value and 2.3% in the 50-year spending they could fund. T-bills were the mirror image: 2.3% and 14.7%.
Elm’s ranking reproduces from public data. Our spending standard deviations of 2.3%, 6.3% and 14.7% for TIPS, nominal bonds and bills sit close to their 2.8%, 6.7% and 14.2%.
The clean result depends on duration matching. Against a 30-year rather than 50-year spending stream, the same TIPS holding’s spending standard deviation rose to 6.7% in our model, because a 30-year bond has roughly the duration of 50 years of spending. A ladder matches a retiree’s horizon; a single long bond does not.
Taxable TIPS hedge inflation but not a real return. In a 24% bracket, each point of extra inflation costs about a quarter point of after-tax real return, and the hedge works best in a tax-deferred account.
A TIPS yield is an informative, imperfect real rate. It includes a liquidity premium that Fed researchers estimate reached up to 300 basis points in 2007 and 2008.

Frequently Asked Questions

Are TIPS risk-free?

No asset is risk-free for every purpose. A TIPS held to maturity pays a known amount of CPI-adjusted dollars on a known date, backed by the U.S. Treasury, which makes it close to risk-free for a spending need of that size and date. Its market price can swing widely before maturity, and it does not protect against inflation in your own spending that runs above CPI-U.

Why did TIPS lose money in 2022 when inflation was high?

In the twelve months to September 2022, the 30-year real yield rose from below zero to about 1.7%, and the price drop from that rise was larger than that year’s inflation adjustment. The longer the bond, the larger the drop. A bond held to maturity still pays its inflation-adjusted principal. Which inflation are you hedging? splits a TIPS fund’s return into those pieces.

Should I hold a TIPS fund or individual TIPS?

For a specific stream of future spending, individual TIPS in a ladder let you match amounts and dates and hold each bond to maturity. A fund keeps a roughly constant duration forever, so it never converges to a known value on a known date. A fund is simpler for a general inflation-linked allocation without a dated liability.

Do T-bills protect retirees from interest-rate risk?

They protect the account balance. They expose the retiree to reinvestment risk: if real yields fall, each dollar buys less future spending. In our replication, the spending a T-bill position could fund fell 19.7% in the year to September 2020 alone.

Is the TIPS yield the real risk-free rate?

It is the closest traded approximation. Federal Reserve research estimates that TIPS yields carry a liquidity premium over the theoretical real rate, large when the market was new and during 2007 and 2008, so the quoted yield overstates the real risk-free rate by an amount that varies over time.

Related Guides

Is Locking In Today’s TIPS Yields Market Timing? for why buying a matched TIPS needs no forecast, and how pensions got better funded in 2022.
How to Build a TIPS Ladder for the mechanics, the 2037 to 2039 maturity gap and a calculator.
Do T-Bills Beat Inflation? for bill returns before and after tax.
Asset-Liability Matching for duration math applied to your own spending plan.