Key takeaways
- A three-fund portfolio held at Vanguard's own 60/40 target weights — 36% total US stock, 24% total international stock, 40% total bond — returned 6.44% a year over the 18 calendar years 2008 to 2025 on the funds' published returns, or 3.91% a year after inflation. One dollar became $3.08, or $2.00 of 2007 purchasing power.
- Measured in nominal terms, 2008 (-21.90%) looks far worse than 2022 (-16.21%). Measured after inflation the two years are 0.63 percentage points apart: -21.88% and -21.25%. The entire difference is that consumer prices fell 0.02% in 2008 and rose 6.41% in 2022.
- The real hole from 2022 was slower to fill. The 2008 loss was made back by the end of 2010; the 2021 real peak wasn't regained until the end of 2025. A more conservative 24/16/60 version was still 1.20% below its 2020 real peak at the end of 2025, five year-ends later.
- Substituting a 10-year Treasury total-return series for the actual bond fund changes the 18-year return by 0.09 percentage points but moves the worst year from 2008 (-21.90%) to 2022 (-18.04%). Which bond series a backtest uses decides which crisis it says was worse.
- Holding 40% of equity abroad cost between 1.32 and 2.02 percentage points a year in every one of the nine ten-year windows inside this sample. Over 2000 to 2009 the MSCI USA index returned 0.05% a year against 3.88% for MSCI World ex USA, so the sample contains one regime, not a law.
Eighteen years, three funds, and 3.91% a year after inflation
Is three funds enough? The three-fund portfolio — the whole US stock market, everything listed outside it, one broad US bond fund — is close to folk wisdom in index investing, which is exactly why it's worth testing rather than repeating.
The short answer from the record is yes, with one specific hole. A portfolio of 36% Vanguard Total Stock Market Index Fund, 24% Total International Stock Index Fund and 40% Total Bond Market Index Fund, rebalanced at each year-end, returned 6.44% a year over the 18 calendar years from 2008 to 2025. After the change in the US consumer price index over the same 18 years, a factor of 1.5419, that is 3.91% a year. A dollar invested at the end of 2007 finished at $3.08, worth $2.00 in 2007 money. Five of the 18 years lost money, both before and after inflation: 2008, 2011, 2015, 2018 and 2022. The international fund is unhedged, so its dollar returns carry the move in the dollar as well as the move in the markets.
Those weights aren't invented. Vanguard's own LifeStrategy Moderate Growth Fund states its target as US stocks 36%, foreign stocks 24%, US fixed income 28% and foreign fixed income 12%. The three-fund version folds that 12% foreign bond slice into the US bond fund and leaves everything else alone.
Every return here is a total return — dividends and coupons reinvested, net of the funds' expenses — taken from the calendar-year figures Vanguard files with the SEC, using the Investor share class throughout. Nothing is a price return. The portfolio series in this piece are built by weighting and compounding those published fund returns. No component series was synthesised — every input is a number a fund or an index provider published.
The rebuilt portfolio can be checked against something published: the LifeStrategy fund itself. Across all 18 years the model's mean absolute yearly difference from it is 0.76 percentage points, and the model's annualised return is 6.44% against the fund's 6.28%. The worst year for the comparison is 2008, where the model is 4.60 points adrift; the fund's underlying holdings have changed over the period, so the early years aren't like-for-like. From 2018 onward the mean absolute difference is 0.32 points and the largest single-year gap is 0.50.
Three funds are broad, and the money inside them isn't evenly spread
The US fund tracks the CRSP US Total Market Index, which Vanguard's prospectus describes as representing "100% of the investable U.S. stock market, as determined by the index provider" and including "large-, mid-, small-, and micro-cap stocks regularly traded on the New York Stock Exchange and Nasdaq". Breadth isn't in question. Weight is.
Because MSCI publishes constituent weights and CRSP does not, the arithmetic below uses MSCI's indexes as of 30 June 2026. MSCI USA held 527 companies covering roughly 85% of US free-float market capitalisation, with 35.53% of its weight in ten of them: NVIDIA at 7.15%, Apple 6.58%, Microsoft 4.08%. Information technology was 37.38% of that index and real estate 1.77%. MSCI ACWI ex USA held 1,934 companies with 17.33% in its top ten, the largest being Taiwan Semiconductor at 5.05%; widen to the ACWI ex USA IMI, 6,018 companies and roughly 99% of the opportunity set outside the US, and the top ten falls to 15.03%.
Pass those weights through a 36/24/40 portfolio and ten American companies account for 12.8% of everything owned, one company for 2.6%, and information technology for about 19% of the whole portfolio including bonds. The actual funds hold the micro-cap tail MSCI USA leaves out, so their figures sit a little below these. The direction is the point: the international fund is the more evenly spread of the two equity sleeves, and it's the smaller one. The look-through arithmetic across overlapping funds reaches the same place from a different set of holdings files.
2008 and 2022 are five points apart in nominal terms and half a point apart in real terms
This is where reading the record carelessly goes wrong. The three test years, as the funds reported them:
| Year | Total US stock | Total international stock | Total bond | Portfolio, nominal | US inflation | Portfolio, real |
|---|---|---|---|---|---|---|
| 2008 | -37.04% | -44.10% | +5.05% | -21.90% | -0.02% | -21.88% |
| 2020 | +20.87% | +11.16% | +7.61% | +13.24% | +1.32% | +11.76% |
| 2022 | -19.60% | -16.05% | -13.25% | -16.21% | +6.41% | -21.25% |
Nominally, 2008 was 5.69 points worse than 2022. In purchasing power the gap is 0.63 points. A real return and a nominal return are different measurements of different things, and putting -21.90% beside -16.21% invites a conclusion the data doesn't carry.
The mechanism sits entirely in the bond sleeve. In 2008 the bond fund gained 5.05% while prices were flat, a real gain of 5.07%, and that is what turned a savage equity year into a bad one. In 2022 the same fund lost 13.25% while prices rose 6.41%, a real loss of 18.47%. The equity sleeves behaved similarly in both years. The defence didn't.
Recovery separates them further. Tracking the portfolio's real value with a running maximum — comparing each year-end against the highest year-end so far, rather than against a peak chosen after the fact — the 2008 drawdown reached 21.88% and was fully made back by the end of 2010. The 2022 drawdown reached 21.25% and the portfolio was still 12.14% below its 2021 real peak at the end of 2023 and 5.90% below at the end of 2024, regaining it only at the end of 2025.
2020 does not appear in the annual record at all
The three-fund portfolio's 2020 was +13.24% nominal and +11.76% real. Nothing in the calendar-year data suggests anything happened.
Something did. Take the equity sleeve alone — 60% MSCI USA, 40% MSCI World ex USA, gross returns in dollars, rebalanced monthly — and apply the same running-maximum rule to month-end values. The fall from the December 2019 peak to the March 2020 trough was 21.06%, and the December 2019 level was back by the end of August 2020. The 2007-09 episode ran to 52.92% and didn't regain its old peak until February 2013. The 2022 drawdown, 25.21% from December 2021 to September 2022, was deeper than 2020's.
Month-end sampling flatters all of these. MSCI's own daily-return calculations put the maximum drawdown since December 1987 at 54.91% for MSCI USA, between 9 October 2007 and 9 March 2009, and 60.58% for MSCI ACWI ex USA, between 31 October 2007 and 9 March 2009. The coarser the grid, the shallower the hole looks — a point that applies to every drawdown statistic, not only this one.
The more conservative the mix, the worse 2022 looks
Shifting the same three funds to 24% US stock, 16% international stock and 60% bonds reverses the ranking of the two crises outright. That portfolio lost 12.90% in real terms in 2008 and 20.32% in real terms in 2022. Its real peak sits at the end of 2020, because 2021 had already cost it 0.67% after inflation. Measured from that peak, the fall ran to 20.86%. At the end of 2025 the portfolio was still 1.20% short of it, five year-ends later.
The pattern runs the length of the ladder. On year-end values the deepest nominal drawdown was 15.22% for a 40/60 mix, 21.90% for 60/40, 30.88% for 80/20 and 39.86% for all-equity. In nominal terms the worst of those year-ends flips from 2008 to 2022 below roughly 46% equity; after inflation the flip happens at about 58%. Holding more bonds moved the worst year rather than removing it.
Change the bond proxy and the worst year changes with it
Run the identical 36/24/40 portfolio with Aswath Damodaran's published annual total return on 10-year US Treasuries in place of the bond fund, and the 18-year annualised return moves from 6.44% to 6.53% — a difference of 0.09 percentage points. The worst year moves from 2008 to 2022. On the Treasury series, 2008 is a 15.88% loss rather than 21.90%, because 10-year Treasuries returned 20.10% that year against the bond fund's 5.05%; and 2022 is an 18.04% loss rather than 16.21%, because the Treasury series lost 17.83% against the fund's 13.25%.
Both series are published; neither was built here. The divergence is a fact about what the fund holds. Vanguard's prospectus describes its target index as measuring "the performance of a wide spectrum of public, investment-grade, taxable bonds in the United States—including government, corporate, and international dollar-denominated bonds, as well as mortgage-backed and asset-backed securities—all with maturities of more than 1 year", with a dollar-weighted average maturity of 8.19 years as of 31 December 2025. Credit exposure hurt it in 2008 and shorter duration helped it in 2022.
A backtest that swaps one for the other matches the headline return to within a tenth of a point and gets "when was the design worst tested" backwards.
International equity cost 1.3 to 2.0 points a year in every window in this sample
Holding equity at 60% and varying only the split, the 18-year annualised return runs from 7.98% with no international exposure to 6.44% at 40% international and 5.66% at 60%. The cost holds inside the sample: across all nine ten-year windows from 2008-2017 to 2016-2025, the 40%-international version trailed the US-only version by between 1.32 and 2.02 percentage points a year. Not one window here paid for the diversification.
That's a fact about 2008 to 2025 and not about diversification. On MSCI's month-end index levels in dollars, MSCI USA returned 0.05% a year from December 2000 to December 2009 while MSCI World ex USA returned 3.88%; from December 2009 to December 2025 the same two indexes returned 14.12% and 7.27%. Across the 187 rolling ten-year windows ending between December 2010 and June 2026, the US index was ahead in 150 and behind in 37, with a maximum shortfall of 3.94 points a year and a maximum lead of 9.18. The test period is one long US regime with the tail of an earlier one attached.
Within crises the international sleeve also behaves inconsistently. In 2008 it fell more than the US fund, -44.10% against -37.04%, so it deepened the loss precisely when a diversifier is supposed to help. In 2022 it fell less, -16.05% against -19.60%. Anyone weighing that trade is really deciding how much home bias to run, which is a separate question with its own evidence.
The objection that three funds omit something that pays, refereed
The standard critique is that the design omits small-cap value, REITs, inflation-linked bonds and commodities. Two of the four aren't omissions.
REITs are inside the US fund: its index covers the whole investable US market, and real estate was 1.77% of MSCI USA at 30 June 2026. Small-cap value is inside it too, at market weight rather than overweight. The critique is really that the market weight is the wrong weight, and over this sample it was not. In Kenneth French's US factor data, the value factor returned -3.76% a year from 2008 to 2025, a cumulative -49.8%, and the size factor -1.24% a year, a cumulative -20.1%. Damodaran's bottom-decile US small-cap series compounded at 6.56% a year over the same 18 years against 10.74% for the total US stock fund. Replacing 10 percentage points of the US sleeve with that small-cap series lowers the portfolio's return from 6.44% to 6.09% and deepens the worst year-end drawdown from 21.90% to 22.66%.
That cuts both ways. Over French's full 1928-2025 record the same two factors returned +3.15% and +1.71% a year. A critic who argues the tilt is worth holding is making a claim about long-run averages, and 18 years can't settle it either way.
The other two are genuine omissions, and one lines up with the hole 2022 exposed. The bond index is investment-grade, taxable and dollar-denominated, so the portfolio carries no inflation-linked or commodity exposure at all. Damodaran's gold series returned 9.60% a year from 2008 to 2025; moving 10 percentage points from the bond sleeve into it raises the portfolio's return to 7.21%, cuts the 2022 loss from 16.21% to 14.83% and leaves the worst year-end drawdown at 21.97%. That is one asset over one period, chosen with hindsight — not evidence the omission is costly in general, but evidence that the specific thing missing is the thing that broke.
The strongest objection to that reading is simply that 2022 is one observation. A single inflation shock in 18 years doesn't establish that the design has a structural weakness rather than an unlucky year, and the portfolio did recover. That objection is fair on the sample size and weak on the mechanism: a bond fund of nominal, investment-grade dollar bonds has no inflation protection by construction, whether or not inflation happens to arrive.
Nothing in this record is a property of the design rather than the period
Splitting the sample in half makes the dependence obvious. From 2008 to 2016 the portfolio returned 4.72% a year nominal and 3.13% real, with the US fund at 7.29%, the international fund at -0.63% and the bond fund at 3.88%. From 2017 to 2025 it returned 8.19% nominal and 4.70% real, with the US fund at 14.30%, international at 8.87% and bonds at 1.82%. Every component ranking changed.
The chart plots the annualised real return of the 36/24/40 portfolio across each of the nine ten-year windows inside the sample. The spread runs from 3.00% a year for 2014-2023 to 6.63% for 2012-2021 — more than double, on the same three funds and the same rule, from start dates nine years apart. Any headline drawn from a single window is a statement about that window.
What this test cannot tell you
It's 18 calendar years of one country's funds in one currency, with no contributions, no withdrawals, no tax and no trading costs beyond the funds' own expenses, which ran 0.14% to 0.17% on the Investor share classes in the 2018 prospectus and are lower now. The annual grid hides everything inside a year, which is why 2020 is invisible in the portfolio numbers.
The composition figures come from MSCI indexes as of 30 June 2026; the return figures come from fund filings through 31 December 2025. Those are different vintages and different index families from the funds' own benchmarks, used side by side here but never inside one calculation. The MSCI equity sleeve behind the monthly drawdowns isn't the three-fund portfolio — it carries no bonds.
Eighteen years contains one severe equity crisis, one fast one and one inflation shock — a small number of observations for a question about design.
What would change the conclusion
A second inflation episode would settle what 2022 only hinted at. If a sleeve of nominal bonds took a second real loss above 15% within a decade, the case that the missing asset is inflation-linked rather than small-cap value would stop resting on one observation.
A decade of international leadership would change the other half. The 2000 to 2009 record shows what that looks like: 0.05% a year for MSCI USA against 3.88% for MSCI World ex USA. Reproduce that, and the 1.32-to-2.02-point drag measured here reverses sign, and every ranking in this piece with it.
A change in what the bond index holds would matter as much as either. The gap between the bond fund's 2008 (+5.05%) and 10-year Treasuries' (+20.10%) is a composition fact, and composition drifts. So does equity concentration: at 35.53% of MSCI USA in ten companies, a compression toward the 17.33% seen in the international index would make the look-through arithmetic here a curiosity.
The figure worth carrying out of all this isn't the 6.44% or the 3.91%. It's the 0.63 percentage points separating 2008 from 2022 once inflation is removed, against the 5.69 points separating them before — and whether the record anyone is relying on was read in the same units throughout. Portfolio trackers including LedgerTouch can mark a portfolio against both, which is a different thing from knowing which one the argument was built on.