Three-month Treasury bills lost ground to inflation more often than they gained on it between January 2000 and September 2026. Rolled for 12 months at a time, they beat the Consumer Price Index in 38.2% of 309 periods, for an average real return of −0.62% before tax. This study pairs the monthly average 3-month Treasury yield with the 12-month change in the CPI-U for every month from January 2000 to September 2026, the latest index, published on October 14, 2026. It measures the real yield two ways, tests every rolling 12-month stretch and every calendar year, and traces the policy eras behind the results. For what bills are and how they are bought, start at the Treasuries hub.
| Measure | Value |
|---|---|
| Latest month | Sep 2026 |
| 3-month Treasury yield (monthly average) | 4.09% |
| CPI inflation over the past 12 months | 3.74% |
| Real yield | 0.34% |
| Share of all 12-month periods since 2000 that beat inflation | 38.2% |
| Longest run of negative real yields | 62 months (Nov 2009 to Dec 2014) |
| Months with a negative real yield since 2000 | 194 of 320 |
Key findings
- Bills lost more often than they won. Of the 309 rolling 12-month periods that began between January 2000 and October 2025, 38.2% beat CPI inflation. The average real return was −0.62%, the worst −7.96% and the best 3.42%.
- Negative real yields were the norm. The month's yield sat below inflation over the previous 12 months in 194 of 320 months, including 62 months in a row from November 2009 to December 2014.
- The era decided the score. Bills beat inflation in 47.9% of periods starting from 2000 to 2007, 17.7% of those from 2008 to 2015, 35.4% from 2016 to 2019, 16.7% from 2020 to 2022 and 97% from 2023 to 2025.
- The 2021–2022 surge was the low point. The monthly real yield reached −7.46% in March 2022, when bills yielded 0.45% and inflation ran at 8.54%, and calendar 2021 returned −6.53% after inflation.
- Bills won the last three calendar years. Real returns were 1.99% in 2023, 2.35% in 2024 and 1.57% in 2025, all before tax.
- The latest margin is thin. In September 2026 the 3-month yield averaged 4.09% against 3.74% inflation, a real yield of 0.34%, after readings of −0.11% in April and −0.54% in May 2026.
How was the real yield measured?
Two ways: each month's yield against inflation over the past 12 months, and the return from rolling bills for 12 months against inflation over those same months.
The first measure is a snapshot, the figure a saver can work out at the time. It divides one plus the month's average yield by one plus the CPI-U's change over the previous 12 months, then subtracts one. The real rate of return guide explains why this exact form beats simple subtraction.
Real yield = (1 + 3-month yield) ÷ (1 + CPI inflation over the past 12 months) − 1
Its weakness is timing. The yield applies to the months ahead, while the inflation figure describes the year behind, and even that arrives late: BLS publishes each month's index in the middle of the following month, so the September 2026 figure came out on October 14, 2026.
The second measure is the outcome, what economists call an ex-post real return. Starting in a given month, the money earns one-twelfth of that month's average yield, moves into the next month at the new yield and keeps going for 12 months. That growth is then divided by the CPI-U's growth over the same 12 months, from the index for the month before the start to the index for the final month. It is what a bill holder actually ended up with, and it can only be known a year after the start. The calendar-year table runs the same test from January to December, against December-to-December inflation.
| Measure | How it is calculated |
|---|---|
| 3-month Treasury yield | Monthly average of the daily 3-month constant maturity yield, quoted on an investment basis (FRED series DGS3MO, from the Federal Reserve's H.15 release) |
| CPI inflation, past 12 months | Change in the CPI-U for all items, US city average, not seasonally adjusted, from the same month a year earlier (FRED series CPIAUCNS, from BLS) |
| Real yield | (1 + yield) ÷ (1 + past 12 months' inflation) − 1, month by month |
| Rolling 12-month real return | Growth from earning each month's yield ÷ 12 for 12 straight months, divided by CPI growth over the same months, minus 1 |
| Share that beat inflation | Rolling periods with a real return above zero, as a share of all periods in the group |
| Calendar-year return | The rolling test from January to December, against December-to-December CPI |
| Eras | Start months grouped as 2000–2007, 2008–2015, 2016–2019, 2020–2022 and 2023–2025 |
The monthly series runs from January 2000 to September 2026, 320 months in all. October 2025 is missing: a lapse in federal appropriations suspended most CPI work, including price collection, from October 1 to November 12, 2025, and BLS published no October 2025 index. The study drops that month and any rolling period that needs its index. For the same reason, BLS has said it will publish no 12-month change for October 2026 when that month's CPI comes out on November 10, 2026.
A constant maturity yield is not the rate from any single auction. Each day Treasury fits a yield curve to closing bid yields on actively traded securities, and the 3-month value is read off that curve at exactly three months. Because it is an investment yield, it lines up with the investment rate in auction results rather than with the lower bank discount rate quoted for bills; the discount rate vs investment rate guide explains the gap.
How has the real yield moved since 2000?
It swung from 3.7% in October 2006 to −7.46% in March 2022 and spent most months below zero, because inflation usually outran the 3-month yield.
- 3-month Treasury yield
- CPI inflation, past 12 months
Wherever the yield line sits above the inflation line, the snapshot real yield is positive. Long spells of that are rare in the record: January 2000 to July 2002, November 2005 to October 2007 and April 2023 to March 2026, with shorter ones in 2009, from late 2018 into 2019 and since June 2026, plus a few scattered months in 2002, 2005, 2015 and 2020. The yield peaked at 6.36% in November 2000 and bottomed at 0.01% in September, November and December 2011. Inflation ranged from −2.1% in July 2009 to 9.06% in June 2022.
The second chart plots the gap itself. Its deepest trough is the 2021–2022 surge, and its longest stretch below zero is the 62 months from November 2009 to December 2014, when the monthly yield never averaged more than 0.16%. The peaks came when inflation dropped quickly while yields held: the 3.7% high of October 2006 followed a fall in 12-month inflation from 4.32% in June 2006 to 1.31% in October, with the yield near 5% throughout.
Did 12 months of rolling bills beat inflation?
In 38.2% of the 309 rolling periods since January 2000, but the share ranged from 16.7% to 97% depending on the era in which a period began.
| Start months | 12-month periods | Share that beat inflation | Average real return | Worst real return | Best real return |
|---|---|---|---|---|---|
| All start months | 309 | 38.2% | -0.6% | -8.0% | 3.4% |
| 2000–2007 | 96.00 | 47.9% | 0.2% | -2.8% | 3.4% |
| 2008–2015 | 96.00 | 17.7% | -1.2% | -3.6% | 2.6% |
| 2016–2019 | 48.00 | 35.4% | -0.5% | -2.3% | 1.3% |
| 2020–2022 | 36.00 | 16.7% | -3.7% | -8.0% | 2.1% |
| 2023–2025 | 33.00 | 97.0% | 1.8% | -0.2% | 3.0% |
Each row groups 12-month periods by their first month, and a period counts as a win when the bills' growth outran the CPI-U over the same 12 months.
The 2000 to 2007 group came closest to an even split, 47.9%, with an average real return of 0.21%. Yields averaged 6% in 2000 and 4.85% in 2006, but only 1.03% to 3.21% a year from 2002 to 2005, below inflation in each of those years. The group's best period, 3.42%, is also the best in the whole sample.
The 2008 to 2015 group covers the first zero-rate era: only 17.7% of its 96 periods beat inflation, and the average was −1.16%. Between the two zero eras, the 2016 to 2019 group won 35.4% of the time as yearly average yields climbed from 0.32% in 2016 to 1.97% in 2018. The 2020 to 2022 group, which runs through the inflation surge, has the lowest share, 16.7%, the lowest average, −3.74%, and the worst single result in the sample, −7.96%.
The 2023 to 2025 group is the reverse. Bills beat inflation in 97% of its 33 periods, all but one, with an average real return of 1.82% and a worst result of −0.23%. It has fewer periods than a full three years would give because each one must end by September 2026, the latest CPI month, and the period that would end on the missing October 2025 index is dropped.
Neighboring periods share 11 of their 12 months, so these shares carry less weight than 309 separate results would; the limitations section below covers this and the choice of eras.
Which calendar years did T-bills beat inflation?
Nine of the 26 years from 2000 to 2025 did: 2000, 2001, 2006, 2007, 2008, 2018, 2023, 2024 and 2025, all before tax.
| Year | Average yield | Inflation | Bill return | Real return | Beat inflation? |
|---|---|---|---|---|---|
| 2000 | 6.0% | 3.4% | 6.2% | 2.7% | Yes |
| 2001 | 3.5% | 1.6% | 3.5% | 1.9% | Yes |
| 2002 | 1.6% | 2.4% | 1.6% | -0.7% | No |
| 2003 | 1.0% | 1.9% | 1.0% | -0.8% | No |
| 2004 | 1.4% | 3.3% | 1.4% | -1.8% | No |
| 2005 | 3.2% | 3.4% | 3.3% | -0.1% | No |
| 2006 | 4.8% | 2.5% | 5.0% | 2.4% | Yes |
| 2007 | 4.5% | 4.1% | 4.6% | 0.5% | Yes |
| 2008 | 1.4% | 0.1% | 1.4% | 1.3% | Yes |
| 2009 | 0.1% | 2.7% | 0.1% | -2.5% | No |
| 2010 | 0.1% | 1.5% | 0.1% | -1.3% | No |
| 2011 | 0.1% | 3.0% | 0.1% | -2.8% | No |
| 2012 | 0.1% | 1.7% | 0.1% | -1.6% | No |
| 2013 | 0.1% | 1.5% | 0.1% | -1.4% | No |
| 2014 | 0.0% | 0.8% | 0.0% | -0.7% | No |
| 2015 | 0.1% | 0.7% | 0.1% | -0.7% | No |
| 2016 | 0.3% | 2.1% | 0.3% | -1.7% | No |
| 2017 | 0.9% | 2.1% | 0.9% | -1.1% | No |
| 2018 | 2.0% | 1.9% | 2.0% | 0.1% | Yes |
| 2019 | 2.1% | 2.3% | 2.1% | -0.2% | No |
| 2020 | 0.4% | 1.4% | 0.4% | -1.0% | No |
| 2021 | 0.0% | 7.0% | 0.0% | -6.5% | No |
| 2022 | 2.1% | 6.5% | 2.1% | -4.1% | No |
| 2023 | 5.3% | 3.4% | 5.4% | 2.0% | Yes |
| 2024 | 5.2% | 2.9% | 5.3% | 2.4% | Yes |
| 2025 | 4.2% | 2.7% | 4.3% | 1.6% | Yes |
Each row rolls bills month by month from January to December and compares the result with the CPI-U from one December to the next. The bill return sits a little above the average yield because each month's interest goes on to earn interest itself: an average yield of 6% in 2000 became a 6.16% return.
Nine straight years, 2009 through 2017, finished behind inflation, with real returns from −0.67% in 2015 to −2.83% in 2011. The year 2008 is the odd one out: bills returned just 1.4%, yet consumer prices rose only 0.09% from December 2007 to December 2008, leaving a real return of 1.31%. The Fed's statement of December 16, 2008 pointed to declines in the prices of energy and other commodities. The year 2018 squeaked through by 0.08%, while 2021 and 2022 were the two worst years in the table, at −6.53% and −4.09%.
Worked example: January's snapshot against the year that followed
The snapshot uses January's average yield and inflation over the 12 months to January. The outcome is the calendar-year real return from the table above.
- 2021: snapshot 1.0008 ÷ 1.014 − 1 = −1.30%; outcome −6.53%. The year finished 5.23 points worse, as inflation climbed from 1.4% in January to 7.04% by December.
- 2022: snapshot 1.0015 ÷ 1.0748 − 1 = −6.82%; outcome 1.021 ÷ 1.0645 − 1 = −4.09%. The year finished 2.73 points better, as the monthly yield rose from 0.15% to 4.36% and inflation eased to 6.45%.
- 2023: snapshot 1.0469 ÷ 1.0641 − 1 = −1.62%; outcome 1.0541 ÷ 1.0335 − 1 = 1.99%. The year finished 3.61 points better, turning an apparent loss into a real gain.
In each case the snapshot missed by several points within a single year, and it missed most when inflation or the Fed changed direction quickly. That gap between what savers could see and what they got is the reason the rolling test exists.
What happened to real yields when the Fed held rates near zero?
They were negative in almost every month, because a yield near zero cannot keep up with rising prices; only months of falling prices produced small positive readings.
The Federal Open Market Committee set a target range for the federal funds rate of 0 to 1/4 percent on December 16, 2008 and kept it there until December 16, 2015, when it raised the range to 1/4 to 1/2 percent. It cut back to 0 to 1/4 percent on March 15, 2020, as the coronavirus outbreak disrupted economic activity, and raised it again on March 16, 2022. Bill yields follow the policy rate closely, as the guide to Fed decisions and bill rates explains, so the 3-month yield sat close to zero in both stretches: 0.01% in three months of 2011 and 0.02% in April and May 2021.
Prices kept rising through most of those years. In the first stretch, 12-month inflation reached 3.87% in September 2011, the month the real yield hit that era's low of −3.71%. With the yield pinned near zero, the real yield was roughly the inflation rate with its sign flipped.
The exceptions came when prices fell. The CPI-U was below its year-earlier level from March to October 2009, and again from January to May 2015 and in September 2015. In those months, yields of a few hundredths of a percent still produced positive real yields, as high as 2.33% in July 2009, when inflation was −2.1% and the yield 0.18%. The same arithmetic gave May 2020 a real yield of 0.01%, the only reading at or above zero in the second stretch's full months at zero, April 2020 to February 2022.
Worked example: $10,000 in rolling bills through 2011
- Bill return for 2011: 0.05%, so $10,000 grew to $10,005.
- CPI-U from December 2010 to December 2011: up 2.96%, so keeping pace needed $10,296.
- Purchasing power at year end: $10,005 ÷ 1.0296 = $9,717.37 in December 2010 dollars.
- Real return: −2.83%, about $283 of buying power gone in a year when the balance rose by $5.
How hard did the 2021–2022 inflation surge hit bill holders?
Harder than any other stretch in the record: the monthly real yield fell to −7.46% in March 2022, and 2020–2022 start months averaged −3.74% over 12 months.
Twelve-month inflation rose from 1.4% in January 2021 to 9.06% in June 2022 on these unadjusted figures. BLS rounded the June reading to 9.1% and called it the largest 12-month increase since the period ending November 1981, with the energy index up 41.6%. The Fed held its range at 0 to 1/4 percent until March 16, 2022, so through all of 2021 the 3-month yield averaged between 0.02% and 0.08% a month while inflation passed 7% by December.
Bill yields then climbed 4.54 points in a year, from 0.15% in January 2022 to 4.69% in January 2023. Inflation started from such a height that the snapshot real yield stayed negative until April 2023, when it turned positive at 0.13%, and it stayed positive in every month with data through March 2026.
On the calendar-year test, 2021 returned −6.53% after inflation and 2022 returned −4.09%. Kept in bills for both years, a balance would have ended with 0.9347 × 0.9591 = 0.8965 of its starting purchasing power, a loss of roughly a tenth.
Securities indexed to the CPI work differently: I bond rates and TIPS principal adjust with the index, which a bill's fixed yield cannot. I bonds vs T-bills compares bills with I bonds, and TIPS explained covers the real yields the market sets in advance.
How do taxes change the results?
They lower them: the study is before tax, federal income tax applies to all bill interest, and states and cities cannot tax it at all.
Federal tax falls on the whole nominal interest, including the part that only makes up for rising prices, so whenever interest is positive the after-tax real return is lower than the figures in these tables. The state side works in bills' favor. Under 31 U.S.C. 3124, obligations of the US government and the interest on them are exempt from taxation by a state or political subdivision, except for certain nondiscriminatory corporate franchise taxes and estate or inheritance taxes. TreasuryDirect sums it up for bills: federal tax is due on the interest, with no state or local taxes.
Bank deposit interest has no such exemption, so the after-tax gap between a bill and a savings account depends on the saver's state. The T-bill tax guide covers reporting and the exemption in detail. Tax rules change, so a tax professional can apply them to a particular situation.
Worked example: 2023 after tax: a bill and a taxable deposit
An illustration, not part of the study: assume a 22% federal rate and a 6% state tax that applies to the deposit but not to the bill.
- Bill return in 2023: 5.41%; after federal tax: 5.41% × (1 − 0.22) = 4.22%.
- Inflation, December 2022 to December 2023: 3.35%.
- The bill's after-tax real return: 1.0422 ÷ 1.0335 − 1 = 0.84%, against 1.99% before tax.
- A deposit paying the same 5.41%, taxed by both: 5.41% × (1 − 0.22 − 0.06) = 3.90%, a real return of 1.0390 ÷ 1.0335 − 1 = 0.53%.
The example ignores the federal deduction for state income tax and any difference in compounding.
What can this study not tell you?
It cannot say what bills will earn next, and its simplifications mean its figures will not match any one investor's statement exactly.
- Averages, not auctions. A real 13-week bill locks in one auction's rate for three months. The study credits each month with that month's average constant maturity yield, which smooths out moves within the month and assumes money can roll every month.
- The snapshot looks backward. It pairs a yield for the months ahead with inflation from the year behind, which is why it missed 2021 and 2023 by several points.
- One price index. The CPI-U averages the spending of urban consumers. A household that spends more on rent, energy or medical care can face higher or lower inflation, and unadjusted figures suit 12-month comparisons, not month-to-month ones.
- Overlapping periods and chosen eras. Rolling periods share most of their months, so the 309 results are not independent, and different era boundaries would give different shares.
- No tax, fees or fund costs. Results are before tax and trading costs; a bill fund would also deduct its expense ratio.
- A gap in the data. October 2025 has no CPI, so the monthly series skips it and one rolling period is dropped.
- One 26-year sample. The span holds two zero-rate eras and one inflation surge. A span reaching back into earlier decades would produce different averages.
What does this record mean for someone holding T-bills?
Bills have reliably returned their dollars but not their purchasing power, and short-term rates relative to inflation set the gap; past performance never predicts future results.
Bills are backed by the full faith and credit of the US government, so the dollars due at maturity are not the open question. Their exposure is to inflation risk, which this study measures, and to reinvestment risk: every 13 weeks a 3-month bill matures and the money rolls at whatever the new rate is, which is how the monthly yield fell from 1.54% in February 2020 to 0.14% by April. A bill sold before maturity can also fetch somewhat more or less than its purchase price plus the interest earned so far. The guide to whether Treasuries are risk-free covers each of these.
The record does not say whether today's margin will widen or shrink. It does suggest questions to ask before comparing cash options. Is the snapshot real yield positive or negative right now? What is left after federal tax? Is the money meant to hold its purchasing power for years, a job that CPI-linked securities are built for? The Federal Reserve held its target range at 3.75% to 4.00% on October 28, 2026; the T-bill rates page shows each term's current yield.
Short-term rates also shape the yield curve. When 3-month yields rose above the 10-year yield in late 2022, the curve inverted, and the yield curve inversion history lists every such episode since 1980.
How to cite or reuse the T-bill inflation data
The dataset is free to reuse under CC BY 4.0: anyone may copy, chart or republish it, with credit and a link to this page.
- Download the CSV, which holds the monthly rows: yield, inflation and real yield for each month. The rolling-period and calendar-year tables can be quoted with the same credit.
- Date every figure, for example: the 3-month Treasury real yield was 0.34% in September 2026, on CPI data published October 14, 2026.
- Keep the caveats with the numbers: results are before tax, October 2025 is missing, and the eras group periods by their start month.
- Suggested credit: Source: Sonic Wallet analysis of Federal Reserve H.15 3-month Treasury constant maturity yields and BLS CPI-U data, via FRED, data through September 2026.
The editorial policy explains how the site's studies are sourced, checked and corrected.