What Makes an Investment Safe? Scoring Cash, Bonds and Stocks Against Your Bills
Since 1928, rolled T-bills never lost dollars over any 10-year span but lost purchasing power in 48% of them. Which asset is safe depends on the bill it pays.
I posted this from my Engineer Investor account on November 12, 2026:
“You can’t really discuss a ‘safe’ asset without specifying the time horizon and whether you mean safety in nominal or real terms. When are your liabilities due? Are they sensitive to inflation? And how flexible are your spending commitments? Safety is fundamentally about matching assets to liabilities.”
Disclosure: I write both Summitward and the Engineer Investor account.
From 1928 through 2025, money rolled through 3-month Treasury bills never ended a 1-, 5-, 10- or 30-year stretch with fewer dollars than it started with. Over the same stretches it ended with less purchasing power between a third and a half of the time. T-bills were a safe home for $30,000 due next year and an unreliable one for spending that rises with prices for decades. Only the bill differed.
Below, T-bills, 10-year Treasuries, the S&P 500 and a 60/40 mix are scored against two kinds of bill over every window since 1928, with and without room for the bill to wait or shrink. Separate guides cover the mechanics that follow: which yield counts as risk-free, duration matching, sizing a bond hedge and building a TIPS ladder.
Three questions that define the bill
- When is it due? A tuition payment next fall, a down payment in three years, or spending every year from 65 to 95.
- What is it measured in? A fixed number of dollars (a mortgage payment, a car you have priced), dollars that rise with consumer prices (most living costs), or another price entirely (a house in your town, private-college tuition).
- Can it bend? Some bills can be postponed, some can be trimmed, and some can do neither.
An asset is safe for a bill when it reliably pays that bill on that date in those units. Price swings in between matter only if you might have to sell before the date.
The scorecard: same assets, two kinds of bill
For every start year from 1928, set aside exactly today’s price of the bill, invest it, and check the balance on the due date. A fixed-dollar bill needs the same number of dollars back. A bill that rises with inflation needs the same purchasing power back, measured with CPI-U. Returns are Damodaran’s annual US series, before fees.1 The table shows the share of windows that fell short.
| Due in | Windows | T-bills | 10-yr Treasury, rolled | S&P 500 | 60/40 |
|---|---|---|---|---|---|
| Fixed-dollar bill: share of windows that ended with fewer dollars | |||||
| 1 year | 98 | 0% | 20% | 27% | 21% |
| 5 years | 94 | 0% | 5% | 12% | 4% |
| 10 years | 89 | 0% | 0% | 6% | 0% |
| 20 years | 79 | 0% | 0% | 0% | 0% |
| Bill that rises with inflation: share of windows that ended with less purchasing power | |||||
| 1 year | 98 | 41% | 43% | 32% | 28% |
| 5 years | 94 | 47% | 39% | 23% | 12% |
| 10 years | 89 | 48% | 37% | 12% | 11% |
| 20 years | 79 | 35% | 33% | 0% | 1% |
| 30 years | 69 | 35% | 33% | 0% | 0% |
Rolling windows by start year, 1928-2025, US data before fees and taxes. 60/40 is the S&P 500 and the rolled 10-year Treasury, rebalanced yearly. The windows overlap: 79 twenty-year windows contain only about four separate twenty-year periods. Source: Damodaran; our calculation.12
For a fixed-dollar bill, cash was safe at every horizon. T-bills never returned fewer dollars than they started with. Stocks did in 27% of 1-year windows and 12% of 5-year windows, and the losses were large: money invested at the end of 1929 had 62% fewer dollars three years later. A down payment you have priced at $80,000 for 2028 is a fixed-dollar bill, and it belongs in T-bills, a CD or a Treasury maturing that year.
For a bill that rises with prices, the ranking flips. T-bills lost purchasing power in 48% of 10-year windows and in 35% of 20- and 30-year windows. The S&P 500 lost purchasing power in 12% of 10-year windows and never over 20 years in this US sample. If T-bill interest is taxed at 24% each year, T-bills fell short in 57% of 10-year windows and 75% of 30-year windows. Do T-Bills Beat Inflation? has the year-by-year record and an explorer for your own tax rate.
A Treasury held to maturity pays its promised dollars
The rolled 10-year Treasury in the table sells its bond every year and buys a new 10-year one, so it always carries ten years of rate risk. A household that buys one 10-year Treasury and holds it to the due date faces a different risk. Damodaran’s workbook carries the 10-year yield at every year-end since 1927, so we can score that household too, treating the bond as a zero-coupon bond bought at the prevailing yield.1
- In dollars it never fell short: in all 89 ten-year windows it paid exactly what it promised.
- In purchasing power it fell short in 25 of the 89 (28%). Most misses came from bonds bought in 1936-1946, 1964-1973 and 2011-2015, by buyers who locked in a yield just before inflation ran ahead of it. A bond bought at the end of 1940 at 2.01% paid 22% more dollars in 1950 and lost 31% of its purchasing power.
- The rolled 10-year Treasury missed in 37% of the same windows.
A nominal Treasury maturing on the due date is the closest thing to a riskless asset for a fixed-dollar bill. For a bill that rises with prices, the matching instrument is a TIPS maturing on the due date, which pays principal indexed to CPI-U. TIPS were first issued in 1997, so they cannot appear in a test that starts in 1928.
Why stocks’ clean 20-year record is weaker evidence than it looks
The 0% in the 20-year stock column comes from one country, the one whose market did best in the 20th century, measured over about four independent 20-year periods. Anarkulova, Cederburg and O’Doherty assembled returns for 39 developed markets and estimate “a 12% chance that a diversified investor with a 30-year investment horizon will lose relative to inflation.”3 Pástor and Stambaugh find that once an investor accounts for uncertainty about expected returns, “stocks are substantially more volatile over long horizons from an investor’s perspective.”4 Bodie showed that the cost of insuring a stock portfolio against earning less than the risk-free rate rises with the horizon, the opposite of what you would expect if stocks became safe over time.5
A longer horizon raises the odds that stocks cover a bill. It does not fix the amount they deliver on the due date. Stocks Are Always Risky covers this evidence in more depth.
Flexibility changes which asset wins
The third question in the post is how flexible the spending is. We reran the test for a 5-year bill that rises with inflation and let it bend two ways: it can wait up to a few extra years for the balance to catch up, or it can be trimmed.
| 5-year bill, rises with inflation | Firm | Can wait 1 year | Can wait 3 years | Can trim 10% | Can trim 20% |
|---|---|---|---|---|---|
| T-bills | 45% | 42% | 38% | 12% | 8% |
| 10-yr Treasury, rolled | 37% | 34% | 26% | 20% | 3% |
| S&P 500 | 24% | 14% | 8% | 15% | 9% |
| 60/40 | 12% | 9% | 5% | 7% | 3% |
Share of the 91 windows starting 1928-2018 (those with three spare years of data) that fell short of the bill’s purchasing power. “Can wait” counts a window as funded if the balance reached the bill at any year-end up to that many years late. Source: Damodaran; our calculation.12
The two kinds of flexibility favor different assets. A bill that can wait three years cut the S&P 500’s miss rate from 24% to 8% and barely moved T-bills (45% to 38%). A bill that can be trimmed 10% cut T-bills from 45% to 12% and stocks only from 24% to 15%. The reason is the shape of the misses. When T-bills fell behind inflation they fell behind by a little, for a long time, so a small cut covers the gap and waiting does not. When stocks fell short they fell short by a lot, then usually recovered, so waiting covers the gap and a small cut does not.
The same holds for a fixed-dollar bill. Over 5 years, stocks ended with fewer dollars in 12% of these windows; given three extra years, none did. Flexibility only counts if it is real. A move can usually slide a year; a tuition bill cannot. A bill that can neither wait nor shrink is the one that needs a matched bond.
Try your own bill
Set the due date, whether the bill is fixed in dollars or rises with inflation, and how far it can bend. Set it to 10 years with no delay to add the Treasury held to maturity.
The underwater chart, replicated
A chart circulating on X shows 10-year Treasuries spending about 46 years below their prior real peak and US stocks about 13. The most authoritative long-run figures come from Dimson, Marsh and Staunton, using daily data on US long government bonds. Bonds peaked in December 1940, fell 67% in real terms, and did not regain that peak until September 1991: “the US bond market’s drawdown, in real terms, lasted for over 50 years.” Their figures for US stocks: a 79% real fall from September 1929, recovered in February 1945.6
Our replication uses year-end data, which misses intra-year peaks such as September 1929. That is why our longest stock spell is shorter than theirs:
| Below a prior real high | Peak | Low point | Recovered | Years |
|---|---|---|---|---|
| 10-yr Treasury, rolled | 1940 | 1981 (-55.5%) | 1991 | 51 |
| T-bills | 1933 | 1951 (-45.8%) | 1997 | 64 |
| T-bills | 2008 | 2022 (-23.0%) | Not by 2025 | 17+ |
| S&P 500 | 1999 | 2008 (-42.2%) | 2013 | 14 |
| S&P 500 | 1972 | 1974 (-48.0%) | 1984 | 12 |
| 60/40 | 1972 | 1974 (-35.2%) | 1985 | 13 |
Real total-return wealth at year-ends, 1928-2025, measured from the prior high to the first year-end at or above it. T-bills’ 1933 high followed four years of deflation, which raised the real value of cash. At the end of 2025 T-bills were still 18.7% below their 2008 real high. Source: Damodaran; our calculation.12
We could not reproduce the chart’s 46 and 13 years from Damodaran’s data or match them to Dimson, Marsh and Staunton’s. The bond spell is real either way. It also describes a fund that never lets its bonds mature. A household that bought a 10-year Treasury at the end of 1940 got every promised dollar back in 1950. What it lost was purchasing power, 31% of it, which is a question about the units of the bill.
When a long bond is safer than cash
For spending that rises with prices and starts years from now, rolling cash is the risky choice. Campbell and Viceira showed that conservative long-horizon investors need to hedge the risk that real interest rates fall, and “long-term inflation-indexed bonds are most suitable for this purpose.”7 Campbell summarized the point for a general audience: “money market investments are not riskless because they must be rolled over,” and “an inflation-indexed long-term bond is actually less risky than cash.”8 Robert Merton made the same argument about retirement savers: measured against the income they need, “the T-bill is actually quite risky.”9
The matched instrument also clears the purchasing-power test on today’s prices. On November 12, 2026, Treasury’s real par yield curve showed 2.64% above inflation at 5 years, 2.86% at 10, 3.11% at 20 and 3.24% at 30.10 A TIPS bought at those yields and held to maturity is contracted to beat CPI-U by that margin before tax. Nominal Treasuries yielded 4.95% at 5 years and 5.10% at 10 the same day.11
When the same bond is riskier than cash
A long bond protects you only if you hold it to the date it was bought for. In 2022, CPI-U rose 6.5% and the Bloomberg U.S. TIPS Index returned -11.85%; the iShares TIPS Bond ETF’s NAV returned -12.13%.112 Real yields rose, and a fund with no maturity date carried that loss to anyone who sold. A household with a fixed bill due in 2023 would have been better off in T-bills.
An inflation hedge also covers only one inflation index over its own horizon. D’Amico and King tested assets against several price indexes and horizons and found that “inflation-linked bonds and swaps are good hedges for headline CPI at horizons matching their maturities,” that they did worse at other horizons and against other indexes, and that “during the inflationary period of 2021-2023, many historical hedging relationships failed.”13 Are TIPS Safer Than They Seem? and Liability-Driven Investing measure how much a TIPS fund or a ladder hedges in practice.
Taxes and whose inflation
Every figure above is before tax except the after-tax T-bill figures. Two tax details change which asset is safe:
- Tax falls on the inflation part of interest. With a tax rate t, inflation π and interest i, the after-tax real return is (1 + i(1 − t)) / (1 + π) − 1. At 5% interest, 3% inflation and a 24% rate, a 1.9% real yield becomes 0.8%.
- TIPS create taxable income before they pay it. The annual inflation adjustment to principal “is taxable in the year in which such adjustments occur even though the inflation adjustments will not be paid until maturity.”14 That is a reason to hold TIPS in an IRA or 401(k) when you can.
The index matters too. TIPS track the non-seasonally-adjusted CPI-U.14 The Bureau of Labor Statistics notes that “because the CPI is a statistical average, it may not reflect your experience.”15 Social Security cost-of-living adjustments use a different index, CPI-W.16 A retiree whose costs run mostly to health care or rent can be fully matched to CPI-U and still fall behind. Which Inflation Are You Hedging? covers the index question.
Where liability matching falls short
- Most household bills are soft. A pension owes a contract. A household chooses most of its spending and can change it. Locking in discretionary spending pays for certainty about an amount you were free to change.
- Matching costs expected return. In a lifecycle model built on developed-market history, Anarkulova, Cederburg and O’Doherty find optimal weights that “remain near one-third domestic stocks and two-thirds international stocks regardless of age,” with “no material fixed income allocation,” and estimate that target-date investors need 63% more savings to match it. It is an unpublished working paper and depends on its model, but it is the strongest case against holding bonds for safety.17
- Diversification and matching do different jobs. A rebalanced mix of stocks and bonds lowers the swings in a portfolio’s value. It does not promise any amount on any date. A matched bond promises one amount on one date and does nothing for the rest of the portfolio.
- Some bills have no Treasury match. A down payment that must cover 20% of whatever houses cost in 2029, or tuition that rises faster than CPI, is tied to a price no Treasury tracks. T-bills can lock in the dollars; they cannot track the house price.
- Wages are an asset too. A 35-year-old will pay most of their future bills from paychecks not yet earned. That is why the bond sleeve in the ALM worked example hedges only 4% of a 40-year-old couple’s liabilities, and why that is fine.
What we recommend
- List the bills by date. Next year, the next five years, and spending after you stop working.
- Label each one. Fixed dollars, rises with inflation, or tied to another price.
- Mark which can bend and whether they can wait or be trimmed.
- Fund each bill with its match.
- Firm, fixed-dollar bills within a few years: T-bills, a Treasury money market fund, insured deposits or a Treasury maturing on the date.
- Firm bills that rise with inflation, years out: TIPS maturing on the date, or a TIPS ladder for a stream of them, in a tax-deferred account if possible.
- Bills that can wait several years: diversified stocks or a stock-and-bond mix, sized to how long the bill can slide.
- Near-term bills that can be trimmed: T-bills or short Treasuries. Over 5 years, T-bills fell more than 10% behind inflation in 12% of windows.
Money with no bill attached goes in a globally diversified portfolio sized to your risk capacity. The Merton Share covers how much risk that portfolio can take.
Key takeaways
- An asset is safe relative to a bill: its date, its units (dollars, purchasing power, or another price) and how much it can bend.
- For fixed-dollar bills, cash was safe. T-bills never ended a window from 1928 to 2025 with fewer dollars; stocks did in 27% of 1-year and 12% of 5-year windows.
- For bills that rise with prices, cash was not. T-bills lost purchasing power in 48% of 10-year and 35% of 30-year windows before tax, 75% of 30-year windows after a 24% tax on interest.
- A Treasury held to maturity pays its dollars. In 89 ten-year windows it never missed in dollars and missed in purchasing power 28% of the time. TIPS maturing on the date are the match for inflation-linked bills.
- Waiting helps stocks; trimming helps cash. For a 5-year inflation-linked bill, three years of slack cut stocks’ miss rate from 24% to 8%; a 10% trim cut T-bills’ from 45% to 12%.
- US stocks’ perfect 20-year record is one country’s history. Across 39 developed markets the estimated chance of a 30-year real loss is 12%.
How Summitward helps
Price your firm spending as a TIPS ladder
The Income Floor planner prices your essential spending at the live Treasury real curve, nets Social Security first, and shows how much of your portfolio a matched floor would take.
Open the Income Floor plannerFrequently asked questions
Is cash the safest investment?
For a bill fixed in dollars and due within a few years, yes. T-bills never returned fewer dollars over any window from 1928 to 2025. For spending that rises with prices over many years, rolled T-bills lost purchasing power in about half of all 10-year windows before tax.
What is the risk-free asset?
It depends on the bill. For a fixed payment on a known date, a Treasury maturing on that date. For spending that rises with inflation on a known date, a TIPS maturing on that date. For next month’s bills, a T-bill or an insured deposit. No single asset is risk-free for all three.
Are TIPS risk-free?
Held to maturity, a TIPS pays a contracted return above CPI-U before tax, backed by the US Treasury. Sold before maturity, its price moves with real yields: TIPS index funds lost about 12% in 2022. CPI-U may also differ from your own inflation, and the inflation adjustment is taxed each year in a taxable account.
Are stocks safe in the long run?
Long horizons improved the odds. In US data since 1928 the S&P 500 never lost purchasing power over 20 years, but that is about four independent periods from the best-performing market of the century. Across 39 developed markets, researchers estimate a 12% chance of a real loss over 30 years. Stocks offer higher expected returns. They do not promise an amount on a date.
Is a high-yield savings account safe?
FDIC-insured deposits up to the coverage limit are safe in dollars. Like T-bills, they do not protect purchasing power over long periods, and the interest is taxed every year. Where to Park Cash compares the options.
Why did bonds stay below their real peak for 50 years?
US long bonds peaked in real terms in December 1940. Inflation in the 1940s and again in the 1970s ran far ahead of the yields locked in earlier, and real values did not recover until 1991. That record describes bond indexes that never mature. Someone who held a single Treasury to maturity got every promised dollar, with less purchasing power.
Related guides
- Should I Buy More Stocks? four investors with the same portfolio and different goals, with a goal funding calculator.
- The Risk-Free Rate Is Your Hurdle Rate which Treasury yield to compare every investment against.
- Do T-Bills Beat Inflation? 98 years of bill returns against inflation, before and after tax.
- Liability-Driven Investing for DIY Investors how much of a bond hedge your spending needs, measured by the hedge ratio.
- Asset-Liability Matching duration matching with a worked example.
- How to Build a TIPS Ladder the instrument that matches inflation-linked spending.
- The Merton Share how much risk to take once the safe asset is chosen.
- Retirement Income Floor whether to lock in essential spending before retiring.
Sources
- Damodaran, A. “Historical Returns on Stocks, Bonds and Bills: 1928-Current”. NYU Stern. Annual S&P 500, 3-month T-bill and 10-year Treasury returns, CPI-U inflation (December to December; 6.5% for 2022), and the year-end 10-year yield from the “S&P 500 & Raw Data” sheet. Every 1928-2025 window, spell and flexibility figure is our calculation from these series.
- Summitward Research. what-makes-an-investment-safe. The script and data behind every table on this page.
- Anarkulova, A., Cederburg, S., & O’Doherty, M. S. (2022). “Stocks for the Long Run? Evidence from a Broad Sample of Developed Markets”. Journal of Financial Economics, 143(1), 409-433.
- Pástor, Ľ., & Stambaugh, R. F. (2012). “Are Stocks Really Less Volatile in the Long Run?”. Journal of Finance, 67(2), 431-478.
- Bodie, Z. (1995). “On the Risk of Stocks in the Long Run”. Financial Analysts Journal, 51(3).
- Dimson, E., Marsh, P., & Staunton, M. Credit Suisse Global Investment Returns Yearbook 2011, p. 7. US real bond and stock drawdowns from daily data.
- Campbell, J. Y., & Viceira, L. M. (2001). “Who Should Buy Long-Term Bonds?”. American Economic Review, 91(1), 99-127.
- Campbell, J. Y. (2000). “Strategic Asset Allocation: Portfolio Choice for Long-Term Investors”. NBER Reporter, Fall 2000.
- Merton, R. C. (2014). “The Crisis in Retirement Planning”. Harvard Business Review, July-August 2014.
- U.S. Department of the Treasury. Daily Treasury Par Real Yield Curve Rates, November 12, 2026.
- U.S. Department of the Treasury. Daily Treasury Par Yield Curve Rates, November 12, 2026.
- iShares Trust. iShares TIPS Bond ETF summary prospectus (Form 497K), March 2023. Calendar-year 2022 returns for the fund and the Bloomberg U.S. Treasury Inflation Protected Securities Index (Series-L).
- D’Amico, S., & King, T. B. (2025). “One Asset Does Not Fit All: Inflation Hedging by Index and Horizon”. Journal of Financial Econometrics, 23(3).
- 31 CFR Part 356. Inflation-Protected Securities. TreasuryDirect. Taxation of inflation adjustments and the CPI-U reference index.
- U.S. Bureau of Labor Statistics. “Why the Published Averages Don’t Always Match An Individual’s Inflation Experience”.
- Congressional Research Service. “Social Security: Cost-of-Living Adjustments” (94-803).
- Anarkulova, A., Cederburg, S., & O’Doherty, M. S. “Beyond the Status Quo: A Critical Assessment of Lifecycle Investment Advice”. Working paper, SSRN 4590406 (July 2025 version).
More in Investing & Portfolio
Browse all investing & portfolio guidesGet new guides by email
Evidence-based, no jargon. At most two emails a month. Unsubscribe any time.
Try it in Summitward
See Monte Carlo retirement simulation in action with your own financial data. Free to start, no credit card required.