If you've watched the 10-year Treasury yield over the past three years, "volatile" undersells it. Since August 2023, the 10-year has ranged from 3.63% to 4.98% — a swing of more than 130 basis points from peak to trough, including a 116-basis-point round trip back up in under four months. The daily standard deviation alone tells you this wasn't a slow drift; it was a series of sharp reversals.

Given that kind of chop, you'd expect a rough few years for bond holders. Instead, the Bloomberg US Aggregate Bond Index ETF (AGG) delivered a 14.25% cumulative return, 4.54% annualized, over that same stretch. There was some pain along the way — a maximum drawdown of -4.86%, driven by the yield spike from 3.63% in September 2024 to 4.79% by January 2025 — but the index climbed back above its prior high by about five and a half months after that January 2025 low, and kept compounding from there.

A quick caveat: Bloomberg US Aggregate Bond Index ETF isn't a pure Treasury fund — it also holds corporates, MBS, and agencies — so this is really a story about how Treasury yield levels have tracked broad investment-grade bond returns, not a 1:1 mechanical link. The relationship has held up well, but it's a proxy, not an identity.

Why didn't the volatility do more damage?

Most of a bond's multi-year return doesn't come from where prices happen to be on any given day — it comes from income. AGG's yield-to-worst back in August 2023 was 4.96%. Three years later, the realized annualized return landed at 4.54% — a gap of just 42 basis points (in the investor’s favor). Despite everything the 10-year did in between, the starting yield turned out to be close to the realized outcome over this time period.

Is that a fluke, or does it hold up historically?

I tested it: 204 rolling three-year windows since 2006, stepped monthly, comparing the starting 10-year yield to the AGG’s subsequent three-year annualized return. Across the full sample, the median absolute gap between the two was 1.6 percentage points — a reasonably tight fit, but not perfect.

One period explains most of the miss. The 36 windows starting between January 2019 and December 2021 — when yields sat near zero and core inflation was about to surge to a four-decade high, forcing the Fed into its fastest hiking cycle in a generation — missed by a median of over 4 points, virtually all in the same direction. Strip those out, and the median gap across the remaining 168 windows tightens to 1.3 points, with nearly half landing within a single point of the starting yield.

Gap between realized return and starting yield, 2006–2023

Source: Morningstar Direct (total return); FRED, series DGS10 (yield). Calculations are the author’s.

Inflation regime may matter more than you’d think

I also split the same 204 windows by whether Core CPI was rising, falling, or roughly flat in the six months before each starting date. The pattern is clean: flat-inflation regimes produced the tightest fit — a median absolute gap of 1.3 points, with a small positive bias (returns modestly beat starting yield). Both rising and falling regimes roughly doubled that error, and rising regimes carried an added problem: a -1.29 point average bias, meaning returns tended to fall short of the starting yield when inflation was actively accelerating — unsurprising, since that’s exactly when the Fed is most likely to hike and potentially push bond prices down.

The rising-regime bias isn’t uniform — starting yield matters there too

That -1.29 point average bias in rising-inflation windows is real, but it isn’t evenly spread. Splitting the 37 rising-CPI windows at their median starting yield (2.0%) shows the bias is concentrated almost entirely in the lower half: windows that started at 2.0% or below missed by a mean of -3.18 points, while windows that started above 2.0% actually missed by a mean of +0.70 points — essentially flat to modestly positive, despite inflation still accelerating. The lower-yield group is almost entirely the 2019–2021 cohort discussed above: starting yields left little or no income cushion when the shock hit.

The current 10-year yield, around 4.7% as of Aug 2026, sits well above even the higher-yield half of that historical split (which topped out near 4.0%). Historically, in this data, a larger starting-yield cushion has coincided with a smaller downside bias when inflation was rising — that is a pattern observed in the past, not a guarantee about how any future inflation surprise would play out, but it is a relevant data point if Core CPI were to turn higher again from here.

Where that leaves us

Core CPI has been in a falling regime for most of the past two years, but the most recent readings — May through July 2026 — show the six-month trend essentially flattening out near 2.5%. In prior periods sharing this pattern, the yield-to-return relationship has tended to be tighter, not looser — though that is a historical tendency, not a projection. It’s also a possible young signal: a few months, not the year-plus stretches that defined past “flat” regimes in this data.

A few months of flat CPI prints doesn’t confirm a durable regime any more than a few months of falling prints did back in late 2024, right before inflation reaccelerated. Historical patterns like these may describe what has happened in the past under similar conditions; they are not a prediction of what core CPI, interest rates, or bond returns will do next, and they should not be relied upon as one.

Bottom line

Pulling the study together: the 10-year treasury bond yield has been genuinely volatile over the past three years, yet the Aggregate Bond Index ETF still delivered a solid 4.54% annualized return — landing within 42 basis points of its starting yield. Widening the lens to 20 years and 204 rolling three-year windows, that relationship holds up as a reasonable, though imperfect, anchor — tightest when inflation is flat, noisier when inflation is trending either direction, and worst of all in the specific case of a low starting yield colliding with a rising-inflation shock, as it did in 2019–2021.

Where that leaves us today: Core CPI has recently flattened after two years of decline, and the starting yield on offer (~4.7%) is well above the level that provided little cushion during the last inflation surprise. Both of those facts may point in a constructive direction for the yield-to-return relationship going forward

That’s the case for staying diligent and diversified: diligent in continuing to track how inflation, yields, and duration positioning interact rather than assuming today’s setup persists, and diversified across maturities, sectors, and asset classes so that no single regime shift — inflation, rates, or otherwise — can do outsized damage to a portfolio built on any one historical pattern.