Strategic Asset Allocation is the long-run policy mix, set from long-run capital-market assumptions to meet a stated investor's return requirement and risk tolerance over a multi-year horizon (PW1.01 supplies the objectives-and-constraints machinery, the Investment Policy Statement, that in practice hands you that target; this node assumes it has already been set). Mechanically, an SAA is a single point on the constrained, investable frontier, chosen to match a target level of risk or return, and then maintained through a rebalancing policy whose calendar-versus-threshold mechanics and cost are AA1.02's job, not re-derived here.
Worked example: Bharat Family Office
For a moderate Bharat Family Office mandate targeting roughly 11% annualized volatility, the closest achievable point on the long-only frontier is, verified in Python:
SAA (target σ ≈ 11%): Equity 54.99%, Debt 26.93%, Gold 18.07%, REIT 0.00%; target return 9.84%, achieved σ = 11.00%, Sharpe = 0.4855.
Hold onto this policy mix, because AA1.03 is built entirely on it: once an SAA like this one is set and priced through the appropriate market indices for each sleeve, that mix becomes the benchmark every subsequent decision is judged against. Every departure from these exact weights (a tactical tilt (see tactical asset allocation), a manager's active security selection within a sleeve) is later measured, segment by segment, against what the policy portfolio alone would have returned. That comparison is Brinson's allocation effect, and it literally does not exist without a policy portfolio defined first, here.
Advertisement
How is the portfolio variance worked out by hand?
As a hand cross-check on the Python output, the same discipline M3.02 modeled for two assets, now run on four, expand σp² = Σ_i Σ_j w_i w_j σ_i σ_j ρ_ij term by term for the SAA weights above (values in %²; off-diagonal pairs appear once below but count twice in the sum):
| Pair | Term (one-sided) | Contribution to Σ (doubled if off-diagonal) |
|---|---|---|
| Equity–Equity | 0.5499² × 19² = 109.173 | 109.173 |
| Equity–Debt | 0.5499 × 0.2693 × 19 × 6 × 0.10 = 1.688 | 3.377 |
| Equity–Gold | 0.5499 × 0.1807 × 19 × 15 × (−0.05) = −1.416 | −2.833 |
| Equity–REIT | (REIT weight = 0) | 0.000 |
| Debt–Debt | 0.2693² × 6² = 2.611 | 2.611 |
| Debt–Gold | 0.2693 × 0.1807 × 6 × 15 × 0.15 = 0.657 | 1.314 |
| Debt–REIT / Gold–REIT / REIT–REIT | (REIT weight = 0) | 0.000 |
| Gold–Gold | 0.1807² × 15² = 7.350 | 7.350 |
| Total | 120.993 (%²) → σ = 11.00% |
The hand sum (120.993) matches the Python w′Σw output to five significant figures, the matrix shortcut and the term-by-term arithmetic are the same claim, checked two ways.