Does Portfolio Rebalancing Improve Returns? Calculate Drift, Trades and Costs

Portfolio rebalancing does not guarantee that selling winners and buying laggards will improve returns. Its first job is to correct an allocation that price movements have changed. The same trade can help when relative performance reverses and hurt when a trend persists. A useful decision therefore starts with the exposure being controlled, the exact trade size and the cost—not a promise of a rebalancing bonus.

Calculate the drift before calculating the trade

Take an unlevered portfolio worth 100,000, with 60,000 in equities and 40,000 in bonds. The target is 60/40. If equities gain 25% while the bond holding is unchanged, the positions become 75,000 and 40,000. Equity weight is now 75,000 ÷ 115,000 = 65.22%, not 60% or 85%. An asset's percentage return cannot simply be added to its portfolio weight.

If the target remains unchanged, the desired positions are 69,000 in equities and 46,000 in bonds. Sell 6,000 of equities and buy 6,000 of bonds. Total wealth is still 115,000 before costs: moving the money does not itself create a return. All figures here are hypothetical. The frictionless baseline permits fractional units, uses total returns including distributions and excludes taxes and execution costs until they are introduced.

For each asset, target weight × current portfolio value − current holding value gives the intended trade: positive for a purchase, negative for a sale. Use one valuation time and one reporting currency, with uninvested cash treated consistently. Comparing yesterday's fund NAV with today's intraday stock price can manufacture apparent drift that is really a timing mismatch.

One rebalance, opposite performance outcomes

Now use a separate two-asset example. Start with 100, split equally between A and B. In period one, A rises 20% and B falls 20%, leaving 60 and 40. Compare holding those units unchanged with selling 10 of A and buying 10 of B at that point, restoring 50/50. The second portfolio trades only once, between the two periods.

The last row is an independent path; it does not use the first period from the other rows. There are no contributions, leverage or forecasts. In the reversal example, each asset ends down 4%, because a 20% rise and a 20% fall do not cancel. Changing the units halfway through nevertheless leaves the portfolio at 100. In the persistent-trend example, trimming the continuing winner sacrifices subsequent upside.

This is not a request to forecast the next reversal. A fixed allocation rule deliberately accepts underperformance on some paths in exchange for tighter control of weights. Choosing whether to rebalance after seeing the second-period result turns hindsight into an apparent strategy. A possible rebalancing benefit is not a guaranteed excess return.

Five percentage points is not a five-percent deviation

For a 60% equity target, an absolute band of five percentage points runs from 55% to 65%. A relative tolerance of 5% around that target instead runs from 60% × 95% to 60% × 105%: 57% to 63%. Those policies allow different drift and can generate different trading. A rule that merely says “rebalance at 5%” is incomplete.

Separate monitoring from execution. A monthly review need not require a monthly trade; the rule may trade only when a band is breached on the review date. Daily monitoring also does not establish the ability to execute at that day's closing price. If a backtest uses closing data to identify a trigger and assumes execution at the same close, it must demonstrate that the order could have been submitted with the information then available.

Specify the destination as well as the trigger. Trading back to the target normally requires a larger adjustment than trading just inside the boundary. Stopping at the boundary saves immediate turnover but may lead to another trigger sooner. No frequency is universally best without a drift tolerance and an executable cost model.

How much new cash avoids selling?

Return to the portfolio holding 75,000 in equities and 40,000 in bonds. If all new cash goes into bonds, restoring equities to 60% requires 75,000 ÷ (115,000 + contribution) = 60%. The contribution must be 10,000. Adding only 5,000 leaves equities at 62.5%: it reduces the deviation but does not eliminate it.

The extra 10,000 is not investment profit. A contribution-funded account cannot be judged against an unfunded account by comparing ending balances. Give both strategies identical cash flows or unitize the account to separate flows from performance, as explained in how deposits and withdrawals distort drawdown charts. Distributions can also fund an underweight category, but a small payment cannot magically fill a larger shortfall.

A planned withdrawal changes the denominator too. Do not use pre-withdrawal target amounts and discover afterward that the remaining mix has drifted further. Establish the required cash and net withdrawal first, then calculate the remaining investable allocation. Borrowing merely to force the spreadsheet back to target changes the strategy.

Why a small deviation may not justify another trade

Selling 6,000 and buying 6,000 creates gross traded notional of 12,000, not 6,000. At a hypothetical all-in friction of 0.2% on each side, the simplified cost estimate is 24. That is an assumption, not a quoted market fee. Minimum commissions, bid-ask spreads, market impact, tax lots and whole-unit restrictions can change the calculation.

Costs paid from the portfolio also reduce the final allocation base. Calculating exact 60/40 positions in a free-trading model and then paying the bill does not necessarily leave exact 60/40 weights. Estimate using executable prices, expected charges and available units; reconcile actual fills and remaining cash afterward. Repeated top-up orders to remove an immaterial decimal difference can create more cost than value.

Tax consequences depend on jurisdiction, account type and the cost basis of the units sold. There is no universal tax rate to apply to every reader. Compare strategies on consistent assumptions and show turnover, net results and allocation drift. A backtest that reveals only the best gross ending value can conceal an expensive implementation.

A rebalancing rule that can actually be audited

1. Define the asset categories, target weights, reporting currency and cash treatment. Confirm that the target still fits the funding horizon; mechanically restoring a deteriorating individual stock is not the same as maintaining broad asset-class exposure.
2. Value everything at a consistent time. Record whether bands use percentage points or relative percentages, the review schedule, the destination weight and the execution timing.
3. Incorporate known contributions, distributions and withdrawals before sizing sales and purchases. Keep cash flows separate from returns.
4. Estimate both sides' costs, taxes and dealing constraints. Reconcile actual weights after execution and record any unfilled amount.
5. Compare with an unrebalanced portfolio using the same start and flows. Report net return, maximum drift, turnover and drawdown, and test on data not used to select the rule.

Successful implementation means returning the account to a still-valid allocation at an acceptable cost, not winning every performance comparison. If liquidity, an asset's mandate or the purpose of the money changes, review the target first. Stable weights do not imply stable volatility, correlations or a fixed maximum loss.