Key takeaways
- Across the 98 calendar years from 1928 to 2025, a 60/40 portfolio left completely alone finished the year a median 3.2 percentage points away from its target — 60/40 became roughly 63/37, or 57/43.
- Drift is lopsided. Equity weight rose in 63 of those 98 years, but the biggest single move was downward: 2008 took a 60/40 to 44.2% equities, while the strongest upward year in almost a century, 1954, only reached 68.9%.
- 2022 breaks the intuition entirely. The S&P 500 returned -18.04% and 10-year Treasuries -17.83%, and the year's drift was 0.06 percentage points. A portfolio can fall 18% and still sit exactly on target.
- Left untouched from the start of 2020, the same 60/40 ended 2025 at 78.2% equities. Vanguard's figure for a global 60/40 running from December 1989 to December 2021 is 80%.
- In Vanguard's 1926-2021 data, a monthly-rebalanced 60/40 had a worst calendar year of -26.6% and an 80/20 had -34.9%. The drifted portfolio's extra 0.9 points of annualised return came with 8.3 more points of worst-year loss.
The median year moves you three points. The worst year moved fifteen.
Set a portfolio to 60% stocks and 40% bonds, then never touch it again. How far off-target is it 12 months later?
Over the 98 calendar years from 1928 to 2025, the median answer is 3.2 percentage points. That is a 60/40 turning into about a 63/37. In 10 of those years the portfolio moved less than a single point. In 28 of them it moved more than 5 points, and in three it moved more than 10.
The kind of year that moves you most isn't the kind most people expect. It isn't the boom. The largest one-year drift in the whole record belongs to 2008, when the S&P 500 returned -36.55% while 10-year Treasuries returned +20.10%, and a 60/40 ended the year at 44.2% equities — 15.8 points below target, and pointed the wrong way for what came next.
How these numbers were calculated, and from what
Every drift figure here's computed rather than quoted, so the inputs are worth stating plainly.
Each calendar year starts at exactly 60% stocks and 40% bonds. That year's total return is applied to each sleeve. There are no contributions, no withdrawals, no trades and no tax. The end-of-year equity weight is 60(1+Rs) divided by 60(1+Rs) + 40(1+Rb), and drift is that number minus 60.
Stocks are the S&P 500 including dividends. Bonds are a constant-maturity 10-year US Treasury, coupon plus price change. Both series are annual, pre-tax, US-dollar total returns from Aswath Damodaran's historical returns dataset at NYU Stern, updated January 2026 and running through 2025. Using the same total-return convention on both sides matters more than which convention it is, because drift is driven almost entirely by the gap between the two sleeves — across the 98 years, the correlation between drift and the stock-minus-bond return gap is 0.99.
Two checks on the data. Damodaran's S&P 500 total returns run 1.3 to 2.3 percentage points above the S&P 500 price index published by the St. Louis Fed for each year from 2017 to 2025 — that difference is the dividend, and it lands where it should. And swapping the Treasury sleeve for Damodaran's Baa corporate bond series moves the median 12-month drift from 3.2 points to 2.8, and 2008's record from -15.8 to -10.4. The magnitudes shift; the shape does not.
Drift measures the gap between your two sleeves, not your losses
2022 is the year that makes this concrete. The S&P 500 returned -18.04% and the 10-year Treasury -17.83% — the 10-year yield went from 1.52% at the end of 2021 to 3.88% a year later, which is what a bond sleeve losing a sixth of its value looks like from the inside. It was, on both sleeves at once, one of the worst years a balanced portfolio has had.
The drift was 0.06 percentage points. A 60/40 ended 2022 at 59.94% equities.
Only five years since 1928 have seen both stocks and bonds fall together: 1931, 1941, 1969, 2018 and 2022. Four of the five produced drift of under 3 points. The exception, 1931, is the one where the two sleeves fell by wildly different amounts (-43.84% against -2.56%).
So a drift number is not a damage report. It answers a narrower question — whether the mix you own still resembles the mix you chose — and a portfolio can be badly down and perfectly on target at the same time.
Upward drift is common and gradual; downward drift is rare and violent
Sorting the 98 years by direction produces something genuinely asymmetric.
- 63 years drifted up toward equities, with a median move of +3.5 points. The largest was 1954: +8.9 points, on a 52.56% equity year.
- 35 years drifted down, with a median move of -2.9 points. The largest was 2008: -15.8 points.
Upward drift happens roughly twice as often but is capped by arithmetic. In almost a century, no single year has pushed a 60/40 past 69% equities. Downward drift is rarer and has a much fatter tail: 2008, 1931 (-13.6), 1937 (-11.1) and 2002 (-9.6) all exceed anything the upside has managed.
The mechanism is that the weight depends on the ratio of the two sleeves' growth, and a ratio isn't symmetric. In 2008 the stock-bond return gap was 56.7 points against equities and drift was -15.8. A hypothetical mirror year — stocks up 56.65%, bonds flat — produces a gap of the same size in the other direction and drift of only +10.2. Equal gaps, unequal consequences.
Which cuts both ways, and this is the part worth sitting with. Any single bad year moves you further than any single good year. But good years outnumber bad ones, so left alone across many years, the drift compounds upward.
A 12-month figure hides how fast the move can happen
Annual data answers "how far in a year". It says nothing about the path, and the path can be violent.
Take 2020. Measured start to finish, that year's drift was a mild +1.4 points — a 60/40 ended at 61.4% equities. Inside the year, the S&P 500 price index fell from 3,386.15 on 19 February to 2,237.40 on 23 March, a decline of 33.9% in 33 calendar days. Holding the bond sleeve flat for illustration, that move alone would take a 60/40 to roughly 49.8% equities — a swing of more than 10 points in about five weeks, fully reversed by December.
Vanguard measured the same episode from the other end and got a similar magnitude. In its December 2024 target-date research, a hypothetical 60/40 rebalanced monthly "could have drifted up to 7% from its target allocation" during March 2020, and one rebalanced quarterly "up to 10% from the target". Those are the drifts that survived a rebalancing rule.
An annual drift figure is therefore a floor on how far a portfolio travelled, not a description of where it went.
Six untouched years turned a 60/40 into a 78/22
The chart above traces a single 60/40 portfolio bought at the start of 2020 and never rebalanced, showing its equity weight at each year-end.
2020 ends at 61.4%. 2021 — equities +28.47%, Treasuries -4.42% — jumps it to 68.1%. 2022 leaves it there, at 68.1%, for the reason described above. Then 2023, 2024 and 2025 add another 10 points: 72.1%, 76.7%, 78.2%.
Six years of ordinary decisions — specifically, no decisions — moved the portfolio 18 points. Nothing dramatic happened in any single year; the largest annual drift in the run was 2021's 6.7 points.
This isn't a peculiarity of the 2020s. Vanguard's October 2022 rebalancing paper runs the same exercise on a global multi-asset 60/40 from December 1989 and reports that it "would have had 80% in equities at the end of 2021", with the never-rebalanced weight ranging "between roughly 50% and 80%" across the period. Running the US stock-and-Treasury version over 1990-2021 gives 86.5%, and extending it to end-2025 gives 91.5% — higher than Vanguard's, which is what you would expect from a US-only equity sleeve over a period when US equities beat the rest of the world.
What the drifted portfolio has actually become
A 78/22 is a different instrument from a 60/40, and Vanguard's 2022 paper happens to quantify exactly the comparison that matters.
Over 1926-2021, on a pre-tax basis with monthly rebalancing, its Figure 2 reports a 60/40 returning 8.9% annualised, with a best calendar year of +36.7% and a worst of -26.6%. An 80/20 returned 9.8%, with a best year of +45.4% and a worst of -34.9%.
That is the trade the drifted portfolio has quietly made on the owner's behalf: about 0.9 percentage points a year of extra expected return, against a worst year 8.3 points deeper. The same direction shows up in the annual data used here — using the 1928-2025 standard deviations (19.4% for stocks, 7.9% for Treasuries, correlation 0.02), portfolio volatility rises from 12.1% a year at 60% equities to 15.3% at 78.2%, about a quarter more.
Vanguard's own framing of why this matters is behavioural rather than statistical: "Meaningful deviations during periods of heightened volatility may cause investors to abandon their portfolio because of a misalignment with their risk tolerance, leading them to lock in their losses."
The case against rebalancing at all is stronger than it sounds
Drift is not automatically a problem, and the strongest counter-argument comes from Vanguard's own numbers.
In Getting Back on Track, a 60/40 run from 1 January 1926 to 31 December 2018 under twelve different rebalancing rules — monthly, quarterly and annual checks, each paired with drift bands of 0%, 1%, 5% and 10% — produced tax-adjusted annualised returns of 8.19% to 8.39%. A thirteenth case sits outside that range and outside those rules: the portfolio that was never rebalanced returned 8.74%, the highest figure in the table, and not by a rounding error.
The rest of that row is the answer. Annualised volatility was 14.0% against 11.4%-11.8% for the rebalanced rules. The Sharpe ratio — return per unit of risk — was 0.46, against 0.50 or 0.51 for every other strategy tested. Average equity allocation was 85%, against 60%-63%. The never-rebalanced portfolio didn't beat the others at the same job; it did a different job.
And the extra return isn't dependable. In the same paper, a 60/40 of global equities and bonds run from 2005 to 2014 ended at $172,170 without rebalancing against $177,082 with a quarterly-monitored 5% threshold rule — the non-rebalanced version trailing "by 5 percentage points after tax over the ten years". Ninety-two years favoured leaving it alone; that particular decade didn't. Whether drift pays depends on which window you happen to live through, which is a weak foundation for a plan.
None of which makes drift bad. It makes it a position: an increase in risk that the market selected rather than the owner. Some people are entirely happy to hold an 85/15 and would have chosen it. The distinction is between arriving there and choosing it. Which rule best holds a chosen line — annual, threshold bands, or something in between — is a separate question, and the evidence on it's covered in our piece on rebalancing frequency and threshold bands.
Drift doesn't predict the next year; it changes how much of it you own
There's a tempting story here — that drift pushes you into equities right before they fall. The data doesn't really support it, and saying so matters more than the story.
In the 21 years where drift exceeded +5 points, the following year's S&P 500 total return averaged 5.83%, against an all-years average of 11.85% — but the median was 10.74%, which is a different picture from the same 21 observations. In the seven years where drift exceeded -5 points, the next year averaged 8.04% with a median of 25.94%. Twenty-one and seven observations cannot carry a timing conclusion, and the mean and median disagree in both directions. There is no signal here worth acting on.
The cost of drift is arithmetic, not prophecy. It shows up as exposure when a bad year does arrive, whenever that is.
Work through 2008 with the actual returns. A portfolio last set to 60/40 at the start of 2003 and never touched entered 2008 at 69.2% equities. In 2008 it returned -19.11%, against -13.89% for a portfolio still at 60/40 — 5.2 points worse. That crash then drove its weight down to 54.3%, and in 2009, when the S&P 500 returned +25.94% and 10-year Treasuries -11.12%, the drifted portfolio made 9.00% against 11.11% at target — 2.1 points worse, in the opposite direction.
Vanguard describes the same round trip in its own data: investors who did not rebalance "would have found themselves overallocated to equities during the correction … precisely when being overexposed would have hurt them the most" and "underexposed to equities during the recovery". The penalty landed twice — 5.2 points on the way down, 2.1 on the way back up — and nobody decided on either.
Where this arithmetic stops applying
Backtests are not forecasts. Everything above is one realised path through history, and it contains an exceptional run for US equities that is the direct cause of the upward skew in drift. A century in which bonds outgrew stocks would produce a mirror image of every number here.
The data is US-only. Damodaran's series is the S&P 500 and US Treasuries; Vanguard's 1989-2021 figure uses global equities and bonds and lands at 80% rather than the 86.5% the US-only version gives over the same window. A portfolio with meaningful non-US exposure drifts less than this piece's headline numbers suggest.
A 60/40 is one portfolio out of many, and the drift arithmetic is specific to it. A 90/10 has far less room to drift upward; an 80/20 with a 5% satellite position in something that moves like crypto behaves nothing like any of this.
Most importantly, the whole calculation assumes a lump sum sitting still. Someone contributing monthly is buying at prevailing weights all year, which pulls the portfolio back toward target during equity rallies and pushes it further during crashes — a materially different path that this dataset can't show. And annual data can't show intra-year drift at all; the 2020 example above uses a price index for the equity leg precisely because the total-return series has no monthly resolution.
Finally, the sources aren't fully independent. Three of the anchors here are Vanguard papers, and Vanguard sells the funds and the advice. Its finding that drift raises risk is unsurprising coming from a firm built on maintaining target allocations, and a reader who discounts it isn't being unreasonable. The counterweight is that the computations in this piece use an outside dataset and reach the same shape.
What would change the conclusion
If stocks and bonds stopped diverging. Drift is the return gap, nothing else. In a decade where the two sleeves move together — 2022 in miniature — a never-touched portfolio would stay close to target on its own, and none of this would be worth measuring.
If the equity risk premium disappeared. The upward skew in drift exists because equities outgrew bonds in 63 of 98 years. Remove that and drift becomes a random walk around the target rather than a one-way ratchet, and the 78.2% result for 2020-2025 becomes a period artefact rather than a tendency.
If the portfolio is fed rather than parked. The 98-year record measures a frozen lump sum. For someone adding cash every month, contributions do part of the work of rebalancing, and the honest answer to "how far does it drift" becomes a different, smaller number that this data can't supply.
The measurement that carries the most information isn't the calendar. It's the distance itself — how far the weights are from the ones that were chosen, and whether that distance is 2 points or 25. LedgerTouch reports that figure continuously and can flag it when it crosses a level you set; a spreadsheet checked once a year reports it too, just less often. What either of them can't decide is whether the number is a problem.