Chapter 2: Asset allocation, the one big decision
A single statistic has sold more financial products than any other sentence in investing: "studies show that 94% of returns come from asset allocation." It appears in seminar slides, robo-advisor pitches, and the openings of a thousand articles, and in that form it is wrong. The study is real, the number is real, and what the number measures is something quite different from what the sales decks claim. This chapter starts by telling the famous story correctly, because the corrected version is the honest case for everything that follows. Asset allocation is your split between the big asset classes, mainly stocks and bonds, and it really is the one big decision. Just not for the reason printed on the slide.
What did the famous study actually find?
In 1986, Gary Brinson and two colleagues studied 91 large US pension funds, then updated the work in 1991. Their finding: a fund's allocation policy explained about 94% of the variance of that fund's returns over time. In plain words, if you want to know why a pension fund had a rough quarter and a smooth year, look at its stock-bond mix, not at which stocks it picked. The study said nothing about 94% of your money, and nothing about 94% of the difference between you and your neighbor. Researchers Roger Ibbotson and Paul Kaplan re-ran the question in 2000 and split it into three honest numbers. Allocation explains roughly 90% of a single fund's ups and downs over time. It explains only about 40% of the differences between one fund and another, because funds with similar mixes are separated by their picks, their timing, and their fees. And it explains about 100% of the average return level across funds, because all the active picking in the market nets out to zero before costs and below zero after them. Three numbers, three different questions, one repeatedly misquoted slide.
Read those three numbers again, because together they are better news than the myth. Your mix decides how your portfolio behaves: how hard it falls in a bad year, how much it sways month to month, what kind of decade it can deliver. Picking does not rescue a wrong mix, and across all investors it cannot, since the picks cancel out and the fees remain. The big decision is the boring one, and it is sitting still, waiting for you to make it on purpose instead of inheriting it from a forum post.
Ninety-nine years on one slider
The calculator below holds Vanguard's published model-allocation data, 1926–2024: ninety-nine calendar years of US stock and bond index returns, blended at every mix from all-bonds to all-stocks. Drag the slider and watch three numbers move.
Set it to 60/40, the most argued-about mix in finance, and the century reads: an average year of 8.8%, a best calendar year of +36.7%, a worst of −26.6%, and $10,000 riding that worst year becomes $7,340. Here is the full table behind the slider.
| Mix (stocks/bonds) | Average year | Best year | Worst year |
|---|---|---|---|
| 0/100 | 5.0% | +32.6% | −13.1% |
| 20/80 | 6.4% | +29.8% | −14.4% |
| 40/60 | 7.7% | +27.9% | −18.4% |
| 60/40 | 8.8% | +36.7% | −26.6% |
| 80/20 | 9.7% | +45.4% | −34.9% |
| 100/0 | 10.5% | +54.2% | −43.1% |
Source: Vanguard's published model-allocation data, 1926–2024. Vanguard no longer names which calendar years produced the extremes, so this guide does not either; the columns are the records, not predictions.
What the split actually controls
Run your eye down the average column. From all-bonds to all-stocks it climbs from 5.0% to 10.5%, a spread of 5.5 points across the entire dial. Now run down the worst column: from −13.1% to −43.1%, a spread of 30 points. The slider you thought was a return dial is mostly a depth gauge. Each 20-point step toward stocks buys roughly one extra point of average year and pays for it with several extra points of worst-year depth: moving from 60/40 to 80/20 adds 0.9 to the average and 8.3 to the worst year. The market sells return, and the currency it accepts is the size of your bad year.
Percentages anesthetize, so the figure below converts the worst column into dollars on a $100,000 portfolio.
Dev re-derives his 80/20
Dev is 28, a software developer with $140,000 in a tidy three-fund portfolio at 80% stocks and 20% bonds. He is proud of the tidiness and a little embarrassed by the origin story: he picked 80/20 because a forum post said it was right for his age. The post may even have been correct. But a mix you inherited is a mix you will abandon, so Dev re-derives his from the table.
At 80/20, the worst calendar year on record was −34.9%. On his $140,000, that is a $48,860 hole, a fall to about $91,140 on paper. One step down the dial, at 60/40, the worst year of −26.6% would have cost $37,240, leaving about $102,760. One step up, at 100/0, the −43.1% record takes $60,340 and leaves $79,660. Dev sits with the three dollar figures rather than the percentages, because percentages are weather and dollars are furniture. His test is concrete: at $91,140, does he keep his automatic contributions running, or does he start drafting the panicked sell order in his head? His income is stable, the money is 30-plus years from its date, and he concludes he can hold the line. So 80/20 survives, this time on purpose. His house fund gets no vote in this decision, by the way, because it never enters the stock portfolio at all; a dated dollar lives by Chapter 1's rules, in bills.
Dev added one sentence under the mix in his plan file: "80/20 means a $49,000 hole in the worst recorded year, and I stay." He says writing the dollar figure felt different from reading the percentage, the way a gym number feels different from a fitness adjective. When the next deep red year arrives, he will not be discovering his loss for the first time. He will be recognizing it, and recognition is most of what composure is made of.
Why do plans use ranges instead of forecasts?
Because the people best equipped to forecast disagree with each other by miles. Vanguard's capital markets model (the March 31, 2026 run) projects US stocks at 4.9–6.9% annualized over the next 10 years, international stocks at 5.0–7.0%, US aggregate bonds at 4.2–5.2%, and cash at 2.9–3.9%. BlackRock's mid-2026 assumptions put US stocks near 8.5% at the center. These are model assumptions, not promises, and the two houses sit roughly 3 points apart on the single most important asset in your portfolio. Both employ careful researchers; both publish methodologies; both have been wrong before in both directions. That 3-point gap is itself the lesson. If the professionals' central estimates disagree by more than the entire historical spread between a 40/60 and a 100/0 average year, then a plan built on any single forecast is built on sand. The historical table earlier in this chapter describes behavior, how hard each mix can fall and how it swings, and behavior is far more stable than levels. Use the table to choose what you can survive. Use the assumption ranges, clearly labeled as assumptions, only to sanity-check whether a goal is roughly fundable. Never use either as a promise.
The decision rule
Everything in this chapter compresses into one procedure. Take your real balance, the number on your actual screen. Multiply it by each worst-year figure in the table. Find the deepest dollar loss you could absorb without selling, not the one that sounds brave at a dinner party but the one that leaves your automatic contributions untouched. Read your mix off that row, and write it down with the dollar figure beside it. Chapter 9 will pressure-test the choice with a rehearsal tool and the difference between what your life can absorb and what your sleep can absorb. For now, a chosen mix with a known worst case beats an inherited mix with a hidden one.
Choose your mix from the worst-year column, in dollars, on your own balance. If you cannot hold the dollar figure, you will not hold the mix, and the average column will never arrive to pay you.
Where people go wrong
- Quoting the 94% myth at themselves. Believing allocation explains everything leads people to ignore fees and chase clever mixes. The honest numbers say the mix sets your behavior and the market sets your level, minus costs. Both halves matter.
- Choosing a mix from the average column. The average year is the advertisement; the worst year is the contract. People who shop the left column and ignore the right one become forced sellers at the bottom, which converts a paper worst year into a permanent one.
- Treating a forecast as a plan. A 10-year projection from any single house, however careful, is one model's guess. Two respectable houses are 3 points apart on US stocks right now. Plans hold ranges; only marketing holds certainties.
- Resetting the dial every January. The mix is a standing decision about your dates and your stomach, neither of which changes with the news cycle. Chapter 10 covers the one maintenance habit the mix actually needs.
Line 1. Each goal, its horizon, its bucket (Chapter 1).
Line 2. My target mix: __% stocks / __% bonds, chosen from the worst-year column on my own balance. Dev's version reads: "80/20, knowing the worst recorded year would cut $140,000 to about $91,140, and I stay."
Key takeaways
- The Brinson study found that allocation policy explained about 94% of the variance of a pension fund's returns over time. It never said 94% of returns, and it never compared investors to each other.
- Ibbotson and Kaplan's three honest numbers: allocation explains roughly 90% of a fund's ups and downs over time, about 40% of the differences between funds, and about 100% of the average return level, since active picking nets to zero minus costs.
- On Vanguard's model-allocation data, 1926–2024, average years span 5.5 points across the dial while worst years span 30. The mix is primarily a depth gauge for your bad year. At the 60/40 default, the average year is 8.8%, the best +36.7%, the worst −26.6%, and $10,000 in that worst year becomes $7,340.
- Forward-looking numbers are model assumptions: Vanguard's March 2026 run puts US stocks at 4.9–6.9% over 10 years while BlackRock's mid-2026 center sits near 8.5%. A 3-point disagreement between experts is the argument for planning in ranges.
- Decision rule: convert the worst-year column to dollars on your balance, find the deepest loss you can hold without selling, take that row as your mix, and write it down.
Sources: Vanguard model-allocation data · CFA Institute: setting the record straight on asset allocation · Vanguard return forecasts (VCMM, March 31, 2026 run) · BlackRock mid-2026 capital market assumptions · Finvest Cash & Bonds Guide