Calendar vs threshold rebalancing

A target allocation drifts the moment markets move, and a "policy" that is never enforced is not a policy. Two disciplines compete:

• Calendar rebalancing: trade back to target on a fixed schedule (quarterly, annually) regardless of how far you have drifted.
• Threshold rebalancing: trade back to target only when a holding's weight breaches a stated band (e.g., ±5 percentage points, absolute), however long that takes.

What are the costs and benefits of each policy?

Each has a genuine cost (bid-ask spread plus any brokerage/impact, applied to the amount traded) and a genuine benefit (controlling drift back toward the risk budget the policy was built to run). Calendar rebalancing is simple and auditable, but it can trade on a date when almost nothing has drifted (cost paid for negligible risk-control benefit) and it can also miss a large drift that happens to build up and then partially reverse between two calendar dates. Threshold rebalancing trades only when drift is real, but it requires monitoring every period and can, in principle, let a slow-building drift run for a long stretch if it never quite crosses the band. Worked Example 6 runs both policies over the same eight-quarter return path for the same portfolio and totals the actual cost of each, the concrete number a written policy has to state.

{
  "skill": "Running a fast what-if sweep on rebalancing bands or a risk-parity solve across many candidate portfolios",
  "use": "Ask an AI coding assistant to sweep a range of threshold bands (e.g., 2-10 percentage points) or calendar frequencies against a return history and report total transaction cost and ending value for each, so a band is chosen with evidence instead of a round-number default.",
  "tools": ["Chat/coding assistants for the sweep script", "A pandas loop over the same quarterly-return table this module builds by hand"],
  "prompt": "Given this table of quarterly returns for [assets], a target policy weight for each, and a per-trade cost of 20bp on amount traded, simulate a calendar rebalance (every N quarters) and a threshold rebalance (band of X percentage points) across a range of N and X. Report ending value, number of rebalances, and total cost for each combination.",
  "verify": "Recompute at least one band's total cost by hand from the trade-by-trade ledger (Guardrail #3 Recompute) before trusting a sweep's recommendation - a solver can apply cost to the wrong base (pre- vs post-trade value) or silently double-count a trade, and a smooth-looking cost curve does not prove the arithmetic underneath it is right. Trace every figure back to a specific quarter's ledger line (Guardrail #1 Trace).",
  "diy": "You must be able to hand-verify one full calendar-vs-threshold comparison end to end (this module's rebalancing worked example) before trusting a machine-generated sweep across dozens of parameter combinations - otherwise a plausible-looking chart can hide a broken cost formula.",
  "market": "IN"
}