A portfolio set at 70% equities and 30% bonds does not stay there. Equities rise, the split becomes 78/22, and the portfolio now carries meaningfully more risk than intended. Rebalancing sells some equities and buys bonds to restore the target.
The question is when. The common answer is a date — annually, or quarterly. The better answer, for most people, is when the portfolio has actually drifted enough to matter.
What changed in 2026
- Automated rebalancing became standard. Robo-advisors and many brokerage platforms rebalance on thresholds automatically, which made band-based approaches the practical default for a lot of investors.
- Fractional shares removed a constraint. Precise rebalancing without odd lots made small adjustments feasible where they previously were not worth the awkwardness.
- Tax-aware rebalancing improved. More platforms prioritise selling tax lots with the smallest gains, which reduces the cost of rebalancing in taxable accounts.
- The research stayed where it was. No approach dominates; the differences between reasonable methods are smaller than the difference between having a rule and not.
Bands versus the calendar
|
Calendar |
Bands |
| Trigger |
A date |
Drift beyond a threshold |
| Trades in calm markets |
Yes, regardless |
No |
| Responds to a sharp move |
Only at the next date |
When it happens |
| Trade frequency |
Fixed |
Variable, usually lower |
| Requires monitoring |
No |
Periodic checking |
| Behavioural risk |
Low |
Moderate — the trigger fires when it feels worst |
Calendar rebalancing trades whether or not anything happened. In a flat year it generates transactions for a portfolio that barely moved, incurring costs and possibly tax for no benefit.
Bands ignore small movements and respond when drift becomes material — including mid-year, which is when it matters most. A sharp market move in March is exactly when a portfolio drifts furthest from target, and calendar rebalancing will not act until December.
The catch is behavioural. A band triggers after a large move, which means the trade it demands is buying the asset that just fell. That is the correct action and the uncomfortable one, and it is the point at which people find reasons to wait.
Setting the bands
Relative bands scale better than absolute ones. An absolute band of 5 percentage points is enormous for a 10% position — a 50% relative move before it triggers — and tight for a 60% position. A relative band, such as 25% of the target weight, adapts: a 10% target triggers at 7.5% or 12.5%, a 60% target at 45% or 75%.
A commonly used starting point is 20–25% relative drift, which for most portfolios means rebalancing every year or two rather than several times a year.
Wider bands for volatile, tax-inefficient positions. If rebalancing a position realises large gains, the threshold should be higher, because the trade costs more.
Check quarterly, act on the band. You need to look periodically to know whether a band has been breached. Checking is not trading — most checks result in no action, which is the design working.
Rebalance with cash flow first
The cheapest rebalancing involves no selling at all.
If you are contributing regularly, direct new money to whatever is underweight. Over time this corrects drift without realising a single gain or paying a single spread. In an accumulating portfolio, contributions alone frequently handle normal drift, and explicit rebalancing trades become rare.
The same works in reverse during drawdown: fund withdrawals by selling whatever is overweight, and the withdrawal rebalances for you.
This matters most in taxable accounts, where selling to rebalance realises gains and creates a tax bill for a transaction that generated no income. Two further points there: rebalance inside tax-advantaged accounts where possible, since trades there have no tax consequence, and consider whether dividends should be redirected rather than automatically reinvested into whatever paid them.
Where a taxable sale is unavoidable, pairing it with loss harvesting can offset the gain — see tax loss harvesting.
Common mistakes
- Absolute bands across differently-sized positions. Too tight for large holdings, far too loose for small ones.
- Rebalancing on trivial drift. Costs and tax exceed the benefit.
- Selling in a taxable account when contributions would have done it. An avoidable tax bill.
- Abandoning the rule after a large move. The moment it fires is the moment it is worth following.
- Ignoring the whole picture. Rebalancing one account while the household allocation drifts elsewhere achieves little.
- Confusing rebalancing with strategy change. Rebalancing restores a target; changing the target is a different decision that deserves its own reasoning.
FAQ
How often will bands actually trigger?
For a typical diversified portfolio with 20–25% relative bands, roughly every one to two years, with clusters around volatile periods. Considerably fewer trades than quarterly rebalancing.
Does rebalancing improve returns?
Primarily it controls risk, keeping the portfolio at the allocation you chose. A return benefit exists in some periods and is not the reliable case. Framing it as risk maintenance rather than return enhancement sets the right expectation.
Should I rebalance during a crash?
That is precisely when it matters, and when it is hardest. A portfolio that has drifted heavily toward bonds after an equity fall has less recovery exposure than intended. Deciding your rule in advance is what makes following it possible.
What about within asset classes?
The same logic applies to sub-allocations, with wider bands — the risk difference between slightly different equity weightings is far smaller than between equities and bonds, so tight bands there generate trades for little benefit.
Where to go next
For the target allocation bands maintain, read glidepaths. For offsetting gains when a taxable rebalance is unavoidable, tax loss harvesting, and for which account each holding belongs in, asset location vs asset allocation.
This is general information, not investment advice. Nothing here accounts for your circumstances, tax position, or horizon.