Suppose you decided, calmly and deliberately, to hold seventy percent stocks and thirty percent bonds. Two good years later, without touching anything, you own eighty-five percent stocks — not because you chose more risk, but because the stocks grew. Your portfolio has silently become a different, riskier portfolio than the one you designed. Rebalancing is simply the act of putting it back.
It sounds trivial. It is actually one of the few investment behaviours that systematically forces you to do the emotionally difficult thing: trim what has been rising and add to what has been falling. Buy low, sell high — implemented not by prediction, but by arithmetic.
What rebalancing is protecting
The target allocation you chose encodes your real constraints: when you need the money, how much decline you can watch without panicking, how stable your income is. Drift undermines exactly that. The eighty-five percent version of your seventy percent portfolio will fall harder in a crash than anything you agreed to — and the moment of discovery is always the worst one.
Rebalancing does not raise returns reliably; some years it costs you gains by trimming a winner early. What it controls is risk. Think of it as steering, not acceleration — the point is arriving where you intended, not arriving fastest.
How often is enough
The honest research answer: it barely matters whether you rebalance yearly or use a threshold like five percentage points of drift — both beat never, and more frequent trading mostly adds costs and taxes without benefit. Once a year, on a date you will remember, is a perfectly respectable policy for a lifetime.
There is also a gentler method: rebalance with new money. Instead of selling winners, direct each fresh contribution toward whichever holding is under target. For anyone still in the saving phase, this achieves most of the correction with no selling, no taxes and no willpower.
The discipline is the product
The deepest value of a rebalancing rule is that it replaces decisions with policy. In a crash, the rule says buy more stocks precisely when every instinct screams sell; in a euphoric market, it trims precisely when everyone else is piling in. You decided the ratio once, in a calm moment — the rule executes your calm self's wishes when your anxious self is in charge.
That is the pattern behind most durable financial habits: automate the wisdom, so the moment cannot renegotiate it. Rebalancing is boring by design. In investing, boring is a compliment.
Calendar bands versus tolerance bands
There are two established ways to decide when to act, and they suit different temperaments. Calendar rebalancing acts on a fixed date regardless of how far the portfolio has drifted. Tolerance-band rebalancing acts whenever a component moves beyond a set distance from its target, regardless of the date.
The calendar version is simpler, requires no monitoring, and can act when nothing needs doing or fail to act when something does. The band version responds to what has actually happened and requires you to be watching, which reintroduces the temptation to look frequently.
The hybrid that most institutional practice has settled on captures the strengths of both: check on a fixed schedule, act only if a band has been breached. This means one review per year or half-year, at most one trade, and no monitoring in between. It is close to the minimum viable version of this discipline and it performs nearly as well as anything more elaborate.
How wide the bands should be
Narrow bands mean frequent trading, more transaction costs, more tax events in unsheltered accounts, and a portfolio held very close to target. Wide bands mean rare trading, minimal costs, and a portfolio that spends much of its time meaningfully away from the intended allocation.
The evidence suggests the outcome is fairly insensitive across a wide middle range, which is unusually good news: there is no precision to be lost by choosing a reasonable figure rather than an optimal one. Something in the region of a fifth of the target weight as a trigger is a common and defensible choice, meaning a component targeted at a given proportion is left alone until it drifts a fifth of that away in either direction.
What is worth avoiding is the extremes. Very narrow bands generate costs that reliably exceed any benefit. Very wide ones effectively abandon the discipline, since a portfolio that only rebalances after enormous drift is a portfolio whose risk profile has already changed for years at a time.
Rebalancing with new money instead of trades
For anyone still contributing, there is a version of this that avoids selling anything at all. Direct each new contribution toward whichever component is currently below its target, and the portfolio drifts back toward the intended weights without a single disposal.
This is meaningfully better in unsheltered accounts because it generates no taxable event, and it is better everywhere because it generates no transaction costs beyond the purchase that was happening anyway. For a portfolio in its accumulation phase with regular contributions, it can handle most ordinary drift on its own.
The limitation is scale. Once the portfolio is large relative to annual contributions, new money cannot correct a substantial drift, and actual rebalancing trades become necessary. The crossover happens gradually and it is worth knowing which regime you are in, because someone in the first regime is doing unnecessary work if they are also trading.
The tax dimension in unsheltered accounts
Selling an appreciated holding to rebalance can realise a taxable gain, which is a real cost that the rebalancing benefit has to exceed. This is why the mechanics differ between sheltered and unsheltered accounts and why treating them identically is a mistake.
Several approaches reduce the cost. Doing all rebalancing inside sheltered accounts where possible, and letting the unsheltered portion drift, achieves the overall allocation with no tax consequence. Using contributions and any withdrawals to adjust weights, as above, avoids disposals. Where a disposal is unavoidable, selling the specific holdings with the smallest gains, if your jurisdiction permits identification of lots, reduces the charge.
Rules vary enormously by country and change over time, so nothing here should be taken as guidance about any specific tax system. The general principle that transfers is that rebalancing has a cost in unsheltered accounts which it does not have in sheltered ones, and the decision threshold should reflect that difference rather than ignoring it.
What the evidence says about the benefit
Rebalancing is sometimes promoted as a source of additional return, on the grounds that it systematically sells high and buys low. The research is more equivocal than that framing suggests, and the honest position is that the return effect is small, inconsistent, and dependent on the specific assets and period examined.
Where the evidence is much clearer is on risk. A portfolio left unrebalanced drifts steadily toward whichever component grew fastest, which over a long period means drifting toward higher risk, since the higher-return asset compounds faster. An investor who chose a moderate allocation twenty years ago and never rebalanced is holding a considerably more aggressive portfolio than the one they selected.
So the accurate claim is that rebalancing is risk control rather than return enhancement. It keeps the portfolio at the risk level you chose, which matters most in the years when a decline would be least survivable. Selling it as a return strategy sets up a disappointment; selling it as maintenance sets an expectation it actually meets.
Why it feels wrong every single time
The consistent experience of rebalancing is that it feels like a mistake at the moment of execution. You are selling the thing that has been working and buying the thing that has not, on the basis of an arithmetic rule rather than any view about what happens next. Every instinct argues against it.
That discomfort is not a signal to reconsider; it is a structural feature of a contrarian rule. If rebalancing felt comfortable, it would mean the components had not diverged, which would mean there was nothing to do. The discomfort and the need are the same thing observed from two directions.
This is the strongest argument for making the rule mechanical and deciding it in advance. A rebalancing policy written down in a calm moment and executed without reconsideration works. One that is reconsidered each time, on the merits, in the presence of recent performance, becomes a market view dressed as a maintenance procedure, and market views are precisely what the discipline was designed to remove.
When the target itself should change
Rebalancing restores a portfolio to its target, which raises the separate question of when the target should move. Confusing these two is common and consequential, because it lets a market-driven decision masquerade as a plan revision.
Legitimate reasons to change a target are all internal: a shortening horizon as retirement approaches, a change in income stability, a new dependant, an inheritance that changes the scale of everything, or the discovery through experience that you cannot tolerate the volatility you signed up for. That last one is a genuine and underrated reason, and adjusting after living through a real decline is a reasonable response to new information about yourself.
Illegitimate reasons are all external and all involve the recent past. A strong run, a poor run, a widely predicted event, a compelling argument about the next decade. None of these are reasons to alter a long-term allocation, and each of them will present itself as one. Writing the legitimate list down in advance, and requiring any change to cite an item from it, is a small piece of structure that protects a portfolio from a great deal of well-intentioned damage. As always, none of this is financial advice.
Rebalancing across accounts rather than within them
Most people hold investments in more than one place: a workplace pension, a personal account, perhaps an older scheme from a previous employer. Rebalancing each of these separately to the same allocation is intuitive and it is not the most efficient arrangement.
The portfolio that matters is the aggregate across all of them, and the target allocation applies to that total rather than to each container. This means individual accounts can hold quite lopsided allocations while the whole remains balanced, which opens up an option: place the components with the least favourable tax treatment inside the sheltered accounts, and let the unsheltered account hold whatever suffers least from being taxed.
The prerequisite is a single view of everything, which most people do not have because the accounts are at different providers. A spreadsheet listing every holding and its value across all accounts, updated when you rebalance, supplies it in about half an hour. Without that view, it is not possible to know what the actual allocation is, which means the rebalancing being performed is against a target nobody has measured.
A policy you can write in five lines
The whole of this can be reduced to a short written statement, and writing it down is what converts an intention into something that survives the moment. Five lines suffice: the target weights, the tolerance band that triggers action, the date of the annual check, whether new contributions are used first, and the list of circumstances that would justify changing the targets themselves.
Keeping this document somewhere you will find it in a year is the only maintenance it requires. Its value appears specifically when markets have moved sharply and you are inclined to reconsider everything, at which point a statement written by a calmer version of yourself is considerably more useful than fresh reasoning.
It also makes the annual review take ten minutes rather than an afternoon, because the questions have already been answered and only the arithmetic is left. That is the practical case for a written policy: not that it produces better decisions in principle, but that it removes the need to make most of them again.