Skip to content

Strategic and Tactical Asset Allocation

One-line definition

Strategic asset allocation (SAA) is the long-run "policy portfolio" — asset-class weights set from an investor's return objective, risk tolerance, and long-run capital market expectations; tactical asset allocation (TAA) is a deliberate, usually short-lived, deviation from that policy portfolio to exploit a near-term view.

Vocabulary

  • Policy portfolio: SAA's other name — the weights an investor would hold absent any short-term view, the baseline everything else is measured against.
  • Active risk / active return from TAA: TAA weights minus SAA weights is exactly an active weight in the same sense a stock-picker's active weight is measured against a benchmark — the policy portfolio is the benchmark here.
  • Rebalancing: distinct from TAA — rebalancing trades back toward the policy weights after market moves drift them away; TAA deliberately trades away from policy weights on a view.

Why PMs care

This is the top-down decision that determines most of a multi-asset portfolio's return and risk before any single-security selection happens — the SAA/TAA split is exactly asking "how much of what we do is a long-run structural bet (SAA) versus a shorter-run market-timing bet (TAA)?" Every optimization tool already in this repo is the engine for setting SAA: efficient_frontier and max_sharpe translate long-run capital market expectations into policy weights the same way they'd translate any other expected-return/covariance inputs into an optimal portfolio; black_litterman_posterior is the standard way TAA views get blended with the market-implied equilibrium prior without those views dominating the whole portfolio; risk_parity_weights and scenario_robust_weights are alternative approaches to setting the same policy-level weights when point-estimate expected returns are distrusted. Nothing new needs to be built here — SAA/TAA is a framing for how this repo's existing optimization machinery gets used in practice, not a separate calculation.

Why this isn't a new src/pm function

There's no new formula: SAA is mean-variance optimization (or Black-Litterman, risk parity, etc.) run once, at a long horizon, on asset-class-level inputs; TAA is the same tools run again at a shorter horizon with updated near-term views, then compared back against the SAA weights the same way active_weights compares a portfolio to a benchmark. Building a distinct "asset_allocation" function would just be re-wrapping functions this repo already has under a new name.

Common mistakes

  • treating TAA as "the active part" and SAA as "the passive part" — SAA is itself an active decision (choosing which long-run weights), just a much lower-turnover one than TAA
  • sizing TAA bets without reference to the SAA policy weights they're deviating from — a TAA tilt only means something relative to what the policy portfolio would otherwise hold
  • confusing TAA with rebalancing — rebalancing restores policy weights after drift; TAA deliberately abandons them on a view. A portfolio can do both at the same time for different reasons

Free resources