Rebalancing: The Mechanics and the Trade-Offs

Scope. MarketClue publishes no rebalancing threshold, recommends no frequency, and does not tell any reader when or whether to rebalance. This article explains what drift is, what rebalancing does to the distribution of outcomes, and why the common claim made for it is wrong. All figures use the illustrative teaching parameters fixed on this pillar's hub and are not forecasts. Costs and tax matter materially to this decision in practice; tax is parked to Annex A.

Holdings that rise become a larger share of the total and holdings that fall become a smaller one, so the mix an investor chose becomes, over time, a mix they did not choose.

Rebalancing is the act of selling some of what has grown and buying some of what has not, to return to the intended proportions. That is the whole mechanism. The interesting part is what it does and does not achieve.

How far a portfolio drifts

Starting at half equities and half bonds, and letting each compound at the hub parameters' expected returns with no intervention at all.

Note that this drift happens with no volatility at all — it is produced purely by one asset compounding faster than the other. Real drift is less orderly, because market movements push the mix around on top of this.

What rebalancing actually does

Simulating 150,000 twenty-year paths from a half-and-half start of $10,000, with annual returns drawn at the hub parameters (equities 9.0% and 16.0%, bonds 5.0% and 6.0%, correlation 0.20), comparing annual rebalancing against leaving the portfolio alone entirely.

Worked example — rebalancing reduced the average outcome, and that is the correct result rather than an argument against it. Leaving the portfolio alone produced a mean of $41,322 against $38,700 for annual rebalancing — 6.8% higher — and the arithmetic says it must: the drifting portfolio's expected value is simply half of each asset compounding on its own, and the faster-compounding asset was never trimmed. There is no puzzle here. The drifting portfolio ended holding a median of 64% in equities rather than 50%, so it earned more because it became a different, higher-risk portfolio — and the dispersion confirms it, with a standard deviation of $21,462 against $15,266, 1.41 times as wide. The trade is visible at both ends: drifting produced a higher 95th percentile ($81,692 against $67,211) and a lower 5th percentile ($18,482 against $18,991), and the medians are almost identical. Notice that the drifting portfolio finished ahead on fewer than half the paths — 47.9% — despite its higher mean: the extra average return lives in the right tail, in the paths where equities did very well and the untrimmed holding rode them. So rebalancing did not cost return in exchange for nothing — it declined to take on the extra risk that would have produced the extra return. (Simulated on annual returns from the stated means, standard deviations, and correlation; the means match the closed-form values — $38,697 rebalanced, $41,286 drifting — and the percentiles vary by a few hundred dollars from run to run without changing the picture.)

The claim that does not survive this

Rebalancing is frequently described as adding return — a "rebalancing bonus" from systematically selling high and buying low.

On these parameters it does not. The rebalanced portfolio earned less on average than the one left alone, and the reason is structural: when one asset has a higher expected return than another, refusing to let it grow as a share of the portfolio necessarily gives up expected return. There is no mechanism by which that can be a free gain.

The real case for it, which needs no bonus

The argument that survives is about identity rather than return: rebalancing keeps a portfolio the one that was chosen.

An investor who selected a half-and-half mix for reasons about their own circumstances, and who finds themselves holding around two-thirds in equities twenty years later, is carrying a risk they never agreed to. The drift was not a decision — and its direction is systematic rather than random. Portfolios drift toward the riskier asset, because the riskier asset is the one that grows faster, so a portfolio left alone becomes progressively less like the one that was designed and progressively more exposed at exactly the point where an investor typically has less time to recover.

Stated plainly: not rebalancing is itself an allocation decision. It is made passively, it moves in one direction, and nobody signed off on it.

What the arithmetic cannot settle

Frequency. More frequent rebalancing tracks the target more closely and incurs more cost. Where the balance lies depends on costs, tax and the size of the account, so this portal publishes no frequency and no drift threshold.

Costs. Every rebalance is a set of transactions, and the friction arithmetic in Pillar 29 applies here unchanged.

Tax. Selling what has risen can be a taxable event, which is a first-order consideration for many investors and is parked to Annex A throughout this portal.

And whether a person will do it. Rebalancing requires buying the thing that has just fallen, at the moment it is least appealing — which is why the discipline is easy to describe and hard to sustain, and why Pillar 32 rather than this one deals with that half.