Key takeaways
- Vanguard's LifeStrategy Moderate Growth Fund held four index funds on 31 October 2025: 36.6% US stocks, 24.8% foreign stocks, 26.8% US bonds and 11.1% foreign bonds.
- Those two equity funds look through to 12,261 holdings, and about 14.1% of the whole fund sits in US technology shares alone.
- On annual US returns from 1928 to 2025, 93.7% of that portfolio's variance came from the equity sleeve. Equity volatility was 19.40% against 7.90% for Treasuries.
- Moving 5 percentage points from equities into gold cut modelled volatility from 12.34% to 11.39%. Equities still supplied 92.3% of the variance.
- The stock and Treasury correlation ran from -0.65 to 0.66 depending on the sample. The equity share of variance never fell below 78.8%.
Four index funds, 12,261 holdings, and one source of risk
You buy a multi-asset fund so you don't have to pick anything, and the factsheet tells you it's diversified. Multi-asset fund diversification is sold as a count of holdings, so the fair question is how much independence that count actually buys.
Here's the answer on one of the larger examples. Vanguard's LifeStrategy Moderate Growth Fund reported net assets of $23.5 billion at its financial year end on 31 October 2025. It held four index funds, a money market fund and two small futures positions. Nothing else.
The weights were 36.6% in Vanguard Total Stock Market Index Fund, 24.8% in Total International Stock Index Fund, 26.8% in Total Bond Market II Index Fund and 11.1% in Total International Bond II Index Fund. That works out at 61.4% in shares, 37.9% in bonds and 0.7% in cash and other net assets.
The two equity funds reported 3,525 holdings at 31 December 2025 and 8,736 at 31 October 2025, so the look-through comes to 12,261 positions. Push those weights through 98 years of US asset returns and 93.7% of the portfolio's variance traces to the equity sleeve. The bonds aren't halving your risk. They're capping it.
What the prospectus targets, and what the filing shows
The design is not a secret. The 2026 prospectus says the fund follows "an asset allocation strategy that reflects an allocation of approximately 60% of the Fund's assets to common stocks and 40% to fixed income securities", with stated targets of 36% US stocks, 28% US fixed income, 24% foreign stocks and 12% foreign fixed income. The filed weights sit within 1.2 points of every one of those targets. Total annual operating expenses are 0.10%.
The same document describes the indirect holdings as "a diversified mix of U.S. and foreign large-, mid-, and small-capitalization stocks". As a description of breadth, that's accurate. It's a claim about how many securities you own, not about how many separate things can go wrong.
Look one level deeper and the breadth narrows. At 31 December 2025, Vanguard Total Stock Market Index Fund held 38.4% of net assets in technology across its 3,525 holdings. That fund is 36.6% of the multi-asset fund, so roughly 14.1% of the whole portfolio is US technology shares. One sector accounts for about a seventh of a fund that names 12,261 holdings.
Multi-asset fund diversification is a variance question, not a count of line items
Here's the arithmetic that settles it. A portfolio's variance is built from three things: the weight of each sleeve, the volatility of each sleeve, and the correlation between them. Volatility enters squared. Correlation enters once, and only through the cross term.
On annual returns from 1928 to 2025, the S&P 500 series in Aswath Damodaran's NYU Stern dataset had a standard deviation of 19.40%. The 10-year Treasury series had 7.90%. The correlation between the two was 0.02, which is about as close to independent as two real assets get.
Put 61.4% into the first and 37.9% into the second and the risk splits 93.7% to 6.3%. The bond sleeve holds 37.9% of the money and carries 6.3% of the variance. Being almost uncorrelated doesn't rescue it. A sleeve that moves roughly two fifths as much as equities cannot contribute much variance whatever it correlates with.
That is the same measurement we applied to a built portfolio in the piece on risk contribution. Here it runs on weights a fund actually filed, rather than weights chosen to make a point. The chart above applies it across the Vanguard range.
The cross-asset correlation matrix behind the number
The matrix below is computed from 98 years of annual US returns, 1928 to 2025, in the same dataset. It is the evidence for the claim that a multi-asset fund has fewer independent bets than holdings.
| Annual returns, 1928 to 2025 | Stocks | Small cap | T.Bills | T.Bonds | Baa bonds | Housing | Gold |
|---|---|---|---|---|---|---|---|
| Stocks (S&P 500) | 1.00 | 0.72 | -0.02 | 0.02 | 0.40 | 0.15 | -0.06 |
| US small cap | 0.72 | 1.00 | -0.15 | -0.10 | 0.31 | 0.15 | 0.01 |
| 3-month T.Bills | -0.02 | -0.15 | 1.00 | 0.28 | 0.12 | 0.08 | 0.13 |
| 10-year T.Bonds | 0.02 | -0.10 | 0.28 | 1.00 | 0.66 | -0.11 | -0.01 |
| Baa corporate bonds | 0.40 | 0.31 | 0.12 | 0.66 | 1.00 | -0.04 | 0.02 |
| Residential housing | 0.15 | 0.15 | 0.08 | -0.11 | -0.04 | 1.00 | 0.08 |
| Gold | -0.06 | 0.01 | 0.13 | -0.01 | 0.02 | 0.08 | 1.00 |
| Standard deviation | 19.40 | 37.95 | 3.04 | 7.90 | 7.65 | 6.18 | 21.51 |
Two rows carry most of the meaning. Large and small US shares correlate at 0.72, so splitting an equity sleeve in two buys less than the two line items suggest. And corporate bonds sit at 0.40 with equities, against 0.02 for Treasuries. A bond sleeve built on a broad aggregate index is part credit, and credit is part equity.
That last point has a measurable cost. Re-run the LifeStrategy Moderate Growth weights with the bond sleeve split 60% Treasury and 40% Baa corporate and the equity share of variance moves from 93.7% to 91.7%, on a slightly higher total volatility. The direction is the one nobody advertises.
A 5% diversifier cannot move a number that large
Gold is the cleanest test available in this dataset. Its correlation with equities over 1928 to 2025 was -0.06, and it is genuinely volatile at 21.51% a year, so it has real risk to contribute.
Move 5 percentage points out of the equity sleeve and into gold and modelled volatility falls from 12.34% to 11.39%. That is a real reduction. But equities still supply 92.3% of the variance, against 93.7% before, and the gold sleeve supplies 0.4%. Take another 5 points into housing and volatility reaches 10.50% with equities still on 90.5%.
The reason is mechanical. At a 5% weight, a sleeve's own variance is scaled by 0.0025. It has to be extraordinarily volatile, or extraordinarily negatively correlated, before that matters next to a 61.4% equity block. Small satellites move the total, but they do not move who is in charge of it, which is the same mechanism as the core-satellite portfolio arithmetic.
Across the fund range, the label moves more than the risk
Read across a whole fund family and multi-asset fund diversification looks less like a dial than a switch. The four LifeStrategy funds share a design and differ only in weights. Equity weights on 31 October 2025 ran 21.5%, 41.3%, 61.4% and 81.4%. The equity share of variance ran 31.8%, 74.9%, 93.7% and 99.0%. By the second rung of a four-rung ladder, three quarters of the risk is already equity risk.
Target date funds behave the same way. Vanguard's Target Retirement 2045 Fund held 49.7% US stocks and 33.8% foreign stocks on 30 September 2025, or 83.5% equity, which produces an equity variance share of 99.3%. Its Target Retirement Income Fund is the most defensive in the range: 36.3% US bonds, 18.5% US stocks, 16.2% short-term inflation-protected bonds, 15.4% foreign bonds and 12.9% foreign stocks. Even at 31.4% equities, the same arithmetic leaves 56.2% of the variance in the equity sleeve.
Actively managed multi-asset funds are not structurally different. The Vanguard STAR Fund held ten positions on 31 October 2025, eight of them equity funds, at 44.9% US stock funds and 19.3% international stock funds against 35.8% in a single bond fund. Using corporate bonds as the fixed income proxy, equities supply 88.8% of its variance.
BlackRock's Global Allocation Fund buys securities directly rather than through underlying funds, and reported 3,339 holdings, $17.2 billion of net assets and a 130% portfolio turnover rate in the year to 30 April 2026. Its own reference benchmark is a 60/40: the S&P 500 at 36%, FTSE World ex-US at 24%, five-year Treasuries at 24% and non-dollar government bonds at 16%. The United States accounted for 66.8% of the fund's total investments. Its ten largest holdings came to 18.0% of total investments: nine megacap technology, internet and semiconductor companies at 16.5%, and a gold exchange-traded fund at 1.5%.
The strongest objection: variance share is not what the bonds are for
Here is the serious counter-argument, and it has real force. Nobody holds bonds in a multi-asset fund to supply variance. They hold them to be somewhere else on the bad days, and a variance decomposition is silent about that.
The 2008 numbers make the case. The S&P 500 series fell 36.55% that year while 10-year Treasuries returned 20.10%. Apply the 61.4/37.9/0.7 look-through weights and the modelled loss is 14.81%. The sleeve carrying 6.3% of the variance is what cut it. On any measure a person actually feels, that is diversification working.
The rebuttal is that the hedge is not reliable. In 2022 the equity series fell 18.04% and the Treasury series fell 17.83%, and the modelled portfolio lost 17.82%. Corporate bonds fell 15.23% in the same year. Splitting the sample makes the instability plain: the stock and Treasury correlation was -0.09 from 1928 to 1959, 0.31 from 1960 to 1999, -0.65 from 2000 to 2021, and 0.66 from 2016 to 2025. That is the correlation instability problem, and it moves the hedge, not the risk split.
And across all four of those sub-samples, the equity share of variance stayed between 78.8% and 114.7%. It reads above 100% in the 2000 to 2021 window precisely because the bond sleeve was subtracting variance rather than adding it. The correlation regime decides whether the bonds help on the bad day. It barely touches who owns the risk.
What this data cannot tell you
The return series is US only, and the funds are not. Foreign equities and currency-hedged foreign bonds are mapped onto S&P 500 and 10-year Treasury series here, which is a simplification, and the direction of the error is unknown. What it does not plausibly do is turn a 93.7% equity share into a balanced one.
The frequency is annual, which gives 98 observations and wide error bars on every correlation in the table. Housing is a price index with no rental income in it, so its return is understated and its correlation with equities is measured on price alone. Gold was administratively fixed for much of the period before 1971, which flatters its independence. And a sample is not a forecast: these are the correlations that happened, not the ones on offer.
One figure did survive an independent check. Compounding the Fama/French monthly market factor plus the risk-free rate over the same calendar years gives an annual standard deviation of 19.83% against the 19.40% used here, and a 2008 return of -36.66% against -36.55%. The equity volatility number is not an artefact of one compiler's choices. The rest of the matrix has no such second opinion in this piece.
What would change the conclusion
Two things would, and both are measurable rather than rhetorical. The first is the volatility gap. Equities ran 19.40% a year against 7.90% for Treasuries over the full sample. Narrow that gap and the whole result moves, because the squared term is what produces the 93.7%.
The second is sleeve size. Funding a hypothetical sleeve with the volatility of equities and zero correlation to anything else out of the equity block, a 10% allocation leaves equities on 88.3% of the variance and a 25% allocation leaves 60.5%. It takes roughly 30% before the equity share drops below half. Alternatively, hold the sleeves as they are and cut the equity weight: against 10-year Treasuries, 30% equities gives a 52.5% variance share and 25% gives 40.3%.
So the threshold that governs multi-asset fund diversification is not the number of holdings, the number of asset classes, or the number of underlying funds. It is whether any non-equity sleeve is large enough, and volatile enough, to register. On the filings read here, only the most defensive fund in the set cleared that bar, and it did so by cutting equities to 21.5% rather than by adding anything clever.