Rebalancing Methods Compared: The Mechanics of Each Approach

Once a portfolio needs rebalancing, there is still a second question that the frequency debate skips over: what does the rebalance actually do when it runs. The main methods are a full restore to target weights, a partial restore to the edge of a tolerance band, cashflow rebalancing using new money and income, one-sided buy-only or sell-only variants, and risk-based schemes that equalise contribution to portfolio volatility rather than capital.

They produce meaningfully different turnover, different residual drift and different tax outcomes from the same starting portfolio. This article lays out the mechanics of each and what each one gives up. It does not name a preferred method, because the ranking depends on cost structure, liquidity, mandate constraints and tax position rather than on anything general.

Step zero: what are you rebalancing back to?

Every method needs a target, and the target scheme decides how much work the rebalancing itself has to do.

A market-capitalisation weighted target self-corrects. When a holding’s market value rises, its target weight rises with it, so a cap-weighted portfolio drifts far less and needs far less trading. An equal weighted target does the opposite: it fights price movement continuously, because every gain pushes a position above its fixed target. The behavioural difference between the two is large enough to deserve its own treatment, in equal weight versus market cap weight.

Between those sit conviction weights, constraint-driven weights (sector caps, single-name caps, liquidity caps), and risk-based weights such as inverse volatility or risk parity. How these are set is covered in position sizing methods.

The target scheme is not a detail. A portfolio that appears to rebalance heavily may simply be running a target that is structurally at war with price movement.

Method one: full restore to target

The mechanic is the plainest. Compute current weights, compute target weights, and trade every difference to zero.

Trades run in both directions in the same session, funded by the sells. The result is a portfolio that exactly matches its design on the rebalance date.

Trade-offs:

• It gives the tightest possible adherence, so drift risk between rebalances starts from zero every time.
• It produces the highest turnover of any method, including many small trades that correct trivial deviations.
• Because it always sells, it realises gains and losses on every run, with the tax consequences that follow.
• It is the easiest method to specify, audit and reproduce in a test, which is a real advantage when a rule has to survive scrutiny.

Method two: partial restore to the band edge

Here a tolerance band sits around each target, and a breached position is traded only far enough to come back inside the band, not all the way to the centre.

If a five percent target carries a band of four to six percent and a holding reaches seven percent, this method sells down to six rather than to five.

Trade-offs:

• Turnover falls substantially compared with a full restore, because each trade is smaller and positions inside the band are left alone entirely.
• Residual drift is deliberately tolerated. The portfolio sits at the boundary of what was declared acceptable rather than at the centre of it.
• Whipsaw risk is higher than restoring to target. A position parked at the band edge can breach again on a small move, so some implementations add an inner buffer, trading back part of the way toward the centre instead.
• Bookkeeping is heavier. The rule now has a target, a band, and a stopping point, and each is a parameter that can be over-tuned in a backtest.

Method three: cashflow rebalancing

Rather than selling anything, incoming cash is directed at the most underweight positions. The sources are new subscriptions, dividends, coupon income, and proceeds from any position exited for research reasons.

Trade-offs:

• No forced sales, so no gains realised purely for the sake of weight maintenance.
• One-sided costs only, and no exposure to spread and impact on the sell side.
• It is slow. Correction happens only at the speed cash arrives, so a portfolio with large drift and small inflows will stay drifted for a long time.
• It works poorly in reverse. A book facing redemptions must sell, and the natural mirror image, funding withdrawals from the most overweight positions, is the same idea running the other way.
• It is naturally suited to portfolios with regular contributions and awkward for a fully invested, static pool of capital.

Method four: one-sided rules

Two variants show up in practice.

Buy-only rebalancing never trims a winner. Overweights are left to run, and underweights are topped up when cash allows. This preserves any momentum in the existing book and avoids realising gains, at the cost of allowing concentration to build steadily. The concentration that accumulates this way needs its own monitoring, as described in concentration risk in portfolios.

Sell-only rebalancing trims positions above a hard cap and holds the proceeds in cash or spreads them across the rest. This is common where a single-name or sector cap is a mandate obligation rather than a preference, and the trim is not optional.

Both are partial by design. They control one tail of the weight distribution and leave the other alone.

Method five: risk-based rebalancing

Instead of restoring capital weights, this family restores risk weights. Position sizes are set so that each holding, or each group of holdings, contributes a similar amount to total portfolio volatility. A low-volatility holding therefore carries more capital than a high-volatility one.

Rebalancing then triggers on changes in estimated volatility and correlation, not only on changes in price.

Trade-offs:

• It targets the thing most investors actually care about, which is risk contribution rather than rupees allocated.
• It depends entirely on estimated volatility and correlation, and those estimates are backward looking, unstable, and prone to shifting exactly when markets are stressed. Correlations rising together in a crisis is a well documented pattern, discussed in reading a correlation matrix.
• Turnover can be higher than a price-based rule, because the inputs move even when prices do not do anything dramatic.
• The lookback window used to estimate volatility becomes another parameter, and short windows produce jumpy allocations while long windows react late.

Method six: doing nothing, stated honestly

Buy and hold is a rebalancing method in the sense that it is a deliberate choice with known consequences. Turnover is minimal, costs are minimal, and no gains are realised for weight maintenance.

What it accepts is unbounded drift. Over long periods a buy-and-hold equity book concentrates into whatever worked, both by name and by sector, and its risk profile at the end can be unrecognisable from its risk profile at the start. That is not automatically wrong. It is only wrong when it happens without anyone noticing, which is why the drift measurement discipline matters more here than anywhere else.

Comparing them on the axes that actually differ

Method	Turnover	Residual drift	Realises gains
Full restore to target	Highest	None on the date	Yes, every run
Restore to band edge	Moderate	Bounded by band	Yes, when breached
Cashflow rebalancing	Low	Persistent until cash arrives	Rarely
Buy-only	Low	One-sided, grows over time	No
Risk-based	Variable, often high	Measured in risk, not weight	Yes
Buy and hold	Minimal	Unbounded	No

The costs sitting behind the turnover column, brokerage, securities transaction tax, stamp duty, spread and market impact, are set out in transaction costs in backtests. They are the reason two methods that look similar on a chart of weights can look quite different on a chart of net outcomes.

What comparing methods does not settle

• None of these methods evaluates a holding. Every one of them treats a position as a weight to be adjusted. Whether the business behind that weight still deserves a place in the portfolio is a research question, and no rebalancing rule answers it. A rule that mechanically buys more of a deteriorating company is doing exactly what it was told to do.
• Backtested rankings between methods are fragile. Which method wins in a historical test depends heavily on whether the sample period trended or mean-reverted, on the cost assumptions applied, and on the universe used. A comparison that reverses when costs are doubled was never a real result.
• Estimation error dominates risk-based schemes. The mathematics is precise, the inputs are not, and the precision of the output can create false confidence.
• Liquidity is the binding constraint that models often ignore. A method that requires selling six percent of a thinly traded holding in one session may be impossible to implement at anything like the modelled price.
• Tax outcomes are investor specific. The same method produces different net results for different holding periods and different entity structures, so a general comparison cannot rank methods on after-tax terms.

Method selection is best understood as a set of declared preferences: how much drift is acceptable, how much turnover is affordable, and which side of the book is allowed to run. Written down that way, the choice becomes something an investment committee can review, rather than a habit nobody remembers adopting.

Related reading

• Portfolio metrics explained: the hub covering the full set of portfolio measures and how they connect.
• How often should you rebalance: the frequency question that sits alongside the method question.
• Measuring portfolio drift: how deviation from target is quantified before any method is applied.
• Position sizing methods: how the target weights being restored are decided in the first place.
• Equal weight versus market cap weight: why the weighting scheme decides how much rebalancing work there is to do.

This article is educational. Altys Labs is not a registered research analyst or investment adviser, and nothing here is investment advice or a recommendation to buy, sell, or hold any security.