Rebalancing normally means selling whatever has done well. There's a quieter version that involves no sale at all. Point every new contribution at the holding that's fallen behind, and the portfolio walks back toward its target on its own.
It works. It also stops working, and the point where it stops is arithmetic rather than opinion. It arrives once the portfolio grows past roughly 11 to 17 times what you add each year.
One equation decides it
Take a portfolio worth V with a target equity weight w. Over a year equities return one number and bonds return another. Call the difference the return gap. Now add a contribution C, all of it to whichever side is light.
If equities led, the money goes into bonds, and the amount that lands the portfolio exactly back on target is (1 − w) × gap × V. If bonds led, the money goes into equities, and the required amount is w × gap × V.
For a 60/40 that's 0.4 times the gap when equities are ahead, and 0.6 times the gap when bonds are ahead. A 10-point gap in favour of equities needs a contribution worth 4% of the portfolio. A 10-point gap the other way needs 6%. Nothing else enters it. Not the level of the market, not volatility, not how long you've held.
The asymmetry is worth sitting with. Topping up the larger sleeve costs more per point of gap, because moving the weight of a 60% holding takes more money than moving a 40% one. Contribution rebalancing is therefore cheapest exactly when equities have run hot, and dearest after they've been hit.
How big the gaps actually get
Aswath Damodaran's NYU Stern dataset carries the annual total return on the S&P 500 and on the 10-year US Treasury back to 1928. Across the 97 calendar years to 2024, the median absolute gap between them was 14.5 percentage points. Equities finished ahead in 62 of those years.
Run the median through the formula and the crossover appears. A 14.5-point gap in favour of equities calls for a contribution worth 5.8% of the portfolio. Invert that ratio and contributions can hold a 60/40 on target while the portfolio stays under about 17 times annual savings. When bonds are the winner, the same median gap needs 8.7%, and the ceiling falls to about 11 times.
Gaps aren't gentle, though. The gap exceeded 10 points in 68 of the 97 years, and 20 points in 34 of them. Exact correction is rarer than the median suggests. This is how often a given contribution rate fully restored a 60/40 over 1928 to 2024, on my calculation from the same series.
| Contribution, as % of portfolio value | Years fully corrected, out of 97 |
|---|---|
| 2% | 9 |
| 4% | 25 |
| 5% | 30 |
| 8% | 56 |
| 10% | 67 |
| 15% | 84 |
Partial correction still does most of the job
Landing exactly on target isn't the test. Staying near it is. So I ran a 60/40 through the same Damodaran series with contributions set to a fixed share of portfolio value each year, always directed at the underweight side, never selling anything.
Over 1990 to 2024 the no-contribution version finished at 90.8% equities and averaged 72.0% along the way. At 1% a year it finished at 82.2%. At 2% it finished at 76.1%. At 5% it finished at 64.2%, averaged 60.1% and never exceeded 65.3%. At 10% it finished on 60.0%.
The full 1928 to 2024 run tells the same story. A 5% contribution rate produced an average equity weight of 60.0%, which is the target to a decimal, with a worst overweight of 68.8% in 1955. A 2% rate averaged 65.0% and peaked at 80.8%. Doing nothing averaged 86.0%. Anyone curious about the single-year version of this drift can see how far a 60/40 wanders in twelve months before any of it compounds.
Around 5% a year is where contribution-only rebalancing stops being a partial fix. That's the 17-times threshold seen from the other side.
Vanguard tested a version of this over 89 years
Zilbering, Jaconetti and Kinniry ran the closest published experiment in Vanguard's November 2015 rebalancing paper. Their Figure 8 takes a 50/50 global stock and bond portfolio from 1926 through 2014 and rebalances it using nothing but the dividends and interest the portfolio threw off.
The redirected-income portfolio drifted to an average equity weight of 53.3%, against a 50% target. A portfolio monitored monthly and rebalanced at 5% thresholds sat at 51.2%. A portfolio left completely alone ran at 80.6%. Returns were 8.1%, 8.1% and 8.9% respectively, and volatility was 9.7%, 10.1% and 13.2%. The income-only version did that with zero rebalancing events and zero turnover, against 64 events and 1.6% annual turnover for the threshold rule.
Two caveats belong next to those numbers. Vanguard assumed no costs at all, and there were no new contributions or withdrawals in any column. And Vanguard flags the bigger one itself: the high dividend and interest levels of that 89-year window may not be available in future. Damodaran's S&P 500 series puts the index dividend yield at 1.25% at the end of 2024, against 2.11% at the end of 2015. An income-only rebalancer now has roughly half the ammunition. Vanguard's own remedy, in the same paragraph, is to use contributions and withdrawals instead, because those don't depend on the level of yields.
What the avoided sale is worth
Two costs disappear when nothing is sold. The first is tax. UK capital gains tax on shares runs at 18% within the basic-rate band and 24% above it, with an annual exempt amount of £3,000 from 6 April 2026. Rebalancing a large taxable holding can therefore hand over a fifth of the realised gain before the new allocation earns anything.
Vanguard's 2019 rebalancing guide put a number on the same idea from the other direction. Split a portfolio evenly between taxable and tax-advantaged accounts, run a quarterly 5% threshold rule, and doing the trading inside the sheltered half beat doing it in the taxable half by 44 basis points a year, with no rise in volatility. That study assumes a 30% income tax rate and a 20% long-term capital gains rate, both American, so the figure indicates the size of the effect rather than a UK number.
For many UK savers the tax saving is nil, and it's honest to say so. Nothing inside an ISA attracts capital gains tax at all, and the allowance is £20,000 a year. A pension is the same. If your whole portfolio sits in wrappers, contribution rebalancing buys you no tax advantage whatsoever.
The second cost survives even there. Vanguard's 2022 analysis models bid-ask spreads rather than assuming them fixed, and finds transaction costs can rise tenfold in market turmoil. It also attributes 10% to 20% of annual rebalancing's advantage over monthly rebalancing to lower transaction costs alone. The same paper notes explicitly that steering periodic contributions into underweight assets rebalances a portfolio naturally, and that allowing for cash flows would justify even less frequent trading or wider bands than its own results suggest. That sits alongside the older comparison of annual rebalancing against 5% threshold bands.
The objection: risk drifts upward for years
Critics of contribution-only rebalancing argue that it's a one-way valve. It buys the laggard and never sells the leader, so through a long bull market the equity weight ratchets up. The saver carries more risk than they chose, for years, and discovers it at the worst possible moment.
The 1990s test that claim directly, and the objection largely survives at low contribution rates. Running 1990 to 1999 with no contributions leaves a 60/40 at 79.5% equities by the end of the decade. At 1% a year it's 76.1%. At 2% it's 73.2% — still 13 points over target after ten years, which is a genuinely different portfolio. At 5% the decade ends at 65.1%. At 10% it ends at 60.4%.
Over the full 1928 to 2024 window the 2% saver peaked at 80.8% equities in 1959. That isn't a rounding error, and it isn't brief. So the objection isn't refuted. It's priced. It bites hard below about 4% and mostly dissolves above 5%, which is the same crossover the formula produces.
There's a second half to the objection that gets less attention and has more evidence behind it. Contribution rebalancing is weakest in the opposite direction. After a crash the underweight asset is equities, and topping up equities is the expensive side of the equation. In the 1990 to 2024 run at 5% a year, the 60/40 finished 2008 at 46.9% equities. That's 13 points below target, not above, and it was still only 57.7% a year later. A seller-rebalancer would have bought back to 60% in the first week of January. The contributor took two years to get there, which matters given what rebalancing did during 2008 and 2020.
Where UK savers sit on the curve
HMRC's savings statistics let you check the ratio against real balances. About £103 billion was subscribed to adult ISAs in 2023 to 2024, across roughly 15 million subscribed accounts, so about £6,900 an account. The average adult ISA market value at the end of 2022 to 2023 was £34,044. That's a contribution rate near 20%, comfortably above the 5.8% the median year demands.
The ratio falls with age, as it has to. Under-25s averaged £8,288, the 25 to 34 group £10,556 and the 65-and-over group £64,386. Even at that top figure, £6,900 is 10.7% of the pot.
Those numbers need handling with care. Market values are for 2022 to 2023 while subscriptions are for 2023 to 2024. The £6,900 is per subscribed account rather than per person, and one person can hold several. HMRC builds the age splits from a sample matched to the Survey of Personal Incomes, and warns that breakdowns by age can fluctuate between years. Directionally, though, the typical UK ISA holder is still well inside contribution-rebalancing range. Someone with £250,000 and £6,900 going in each year isn't. That's 2.8%.
When the crossover arrives
The ratio decays on its own, and quickly. Start from nothing, save a fixed amount, earn 5% a year, and the portfolio reaches 12.6 times the annual contribution after ten years, 17.7 times after thirteen, 33.1 times after twenty and 66.4 times after thirty. Put differently, the contribution is 8.0% of the portfolio at year ten, 3.0% at year twenty and 1.5% at year thirty.
Let contributions rise 3% a year with pay and the schedule stretches, but not by much. The contribution is 6.0% of the portfolio at year fifteen, 4.3% at year twenty and 2.1% at year thirty-five.
Thirteen to seventeen years of steady saving, then, and new money on its own no longer holds the line in a median year. For most people that lands decades before retirement, which is also when the gap between stated risk tolerance and actual risk capacity starts to matter. Tracking the ratio is trivial arithmetic, and a drift report in LedgerTouch shows the same thing from the allocation side.
What would change the conclusion
The equation is fixed. Everything empirical here rests on the distribution of equity-versus-bond return gaps, and that distribution can shift.
A long run of small gaps would stretch the ceiling a long way. In 2022 the gap was 0.2 points, because both sides fell together. Several years like that and contributions worth 2% of a portfolio would be plenty. A run of 2013-style years, where the gap hit 41 points, would shrink the ceiling to single-digit multiples.
The two-asset frame also flatters the method. A portfolio with five sleeves drifts on several axes at once, and one contribution can only be pointed at one of them. Correcting a five-way drift with new money needs a bigger contribution than correcting a two-way one.
The direction of cash flow matters too. In retirement the sign flips: withdrawals taken from the overweight asset do the identical job, and a decumulating portfolio can hold its target without selling into a target-driven rule for as long as the withdrawal rate covers the drift.
Finally, the return series is American. Damodaran's numbers are the S&P 500 against the 10-year Treasury. A global equity index paired with global bonds would generate a different gap distribution, and probably a slightly narrower one, which would push the crossover multiple higher. The direction of the finding wouldn't change. The precise multiple would.