Rebalancing

Definition

Rebalancing restores a portfolio to its target weights after market moves drift them — mechanically selling what rose and buying what fell. Perold-Sharpe (1988) is the clarifying frame: rebalancing (constant-mix) is a CONCAVE strategy that profits in oscillating markets and lags in trends, while buy-and-hold and momentum-style strategies are the convex mirror — neither dominates; the choice buys one payoff shape by selling the other. “Rebalancing bonus” claims are the oscillation half of that trade wearing a free-lunch costume.

How it works / structure

• Mechanics (engine-executable): calendar triggers (monthly/quarterly/annual), threshold/band triggers (rebalance when a weight drifts ±X% — fewer, larger trades; documented as more friction-efficient than calendars for most vol levels), and hybrid (calendar-checked bands). Parameters: bands, trade-to- target vs trade-to-band-edge, and the friction model (ms-slippage-friction — rebalancing is a standing friction purchase).
• The Perold-Sharpe geometry: constant-mix sells strength/buys weakness (concave — harvests mean reversion, suffers in sustained trends, and buys all the way down in crashes); buy-and-hold holds the drift (convex-neutral); the payoff-shape choice is the actual decision, prior to any parameter.
• The “bonus” honestly stated: volatility harvesting (rebalancing captures part of σ²/2 when assets oscillate without trending) is real arithmetic CONDITIONAL on mean reversion at the rebalance horizon — it is negative arithmetic under trends; the platform quotes it as conditional, never as free.
• Frictions and taxes: every rebalance realizes gains in taxable accounts (acct-wash-sale family — facts, never advice) and pays spread/impact; band design is a friction- vs-drift optimization with measurable terms.

When it applies

Policy portfolios with fixed targets (port-allocation-frameworks fixed-weight and risk-based frameworks both need a rebalance rule); multi-strategy books (re-allocating across strategies at review horizons — mgmt-scaling is the position-level cousin); vol-targeting overlays (their rebalancing IS the mechanism — risk-volatility-targeting).

Risk profile & failure modes

• Trend bleed: constant-mix’s documented cost — rebalancing out of a multi-year winner surrenders exactly the tail strategy-momentum documents and strategy-buy-and-hold’s evidence rides; band width is the dial between harvesting and bleeding.
• Crash-buying without a floor: mechanically buying a falling asset re-loads risk all the way down — constant- mix has no stop; pairing budgets (risk-max-drawdown-budget) bound what the mechanism may spend.
• Over-frequent friction burn: tight bands + volatile assets = a friction annuity paid to market makers; the trigger design is measurable and replayable, not aesthetic.
• Tax blindness: rebalancing identically in taxable and deferred accounts ignores a first-order cost difference (acct-account-types facts).

Evidence & limits

Perold-Sharpe (1988) is the canonical payoff-shape analysis; volatility-harvesting arithmetic is mathematics with a documented conditionality; band-vs-calendar efficiency comparisons are practitioner-replicated. No rebalancing rule claims return superiority in general — the platform replays the chosen rule against buy-and-hold on the same book and quotes both shapes.

Falsifiable-thesis examples

Illustrations only, not signals:

• “±5% bands will produce fewer rebalance trades and less friction than quarterly calendar rebalancing on this book, with tracking within 1% (band-efficiency thesis)” — falsified by the paired replay.
• “Constant-mix will beat buy-and-hold on this two-asset book over the next 3 years (oscillation-regime thesis)” — falsified by the paired outcome.

Cross-references

• The framework it serves: port-allocation-frameworks
• The payoff-shape mirror: strategy-buy-and-hold, strategy-momentum (convex side)
• The costs: ms-slippage-friction, acct-wash-sale, acct-account-types
• The position-level cousin: mgmt-scaling