Performance Attribution That Doesn't Add Up

First compute the policy benchmark return: 0.60 x 11.2% + 0.40 x 3.0% = 6.72% + 1.20% = 7.92%. The fund returned 9.4%, so excess return versus policy is 1.48%, not 1.3% (the 1.3% figure is versus the 8.1% benchmark, which is a different benchmark). This is the first reconciliation issue: the client is comparing to the wrong benchmark.

Decompose versus policy. Allocation effect: the manager held 65% equity versus 60%, so the equity overweight contributed (0.65 - 0.60) x (11.2% - 7.92%) = 0.05 x 3.28% = +0.16%. The bond underweight contributed (0.35 - 0.40) x (3.0% - 7.92%) = -0.05 x -4.92% = +0.25%. Total allocation +0.41%.

Selection effect: within equity, the tech overweight and energy underweight. If tech was 30% of equity versus 22% benchmark and energy 4% versus 9%, selection within equity is roughly (0.30-0.22) x (18%-11.2%) + (0.04-0.09) x (-4%-11.2%) = 0.08 x 6.8% + (-0.05) x (-15.2%) = 0.54% + 0.76% = +1.30% on the equity sleeve, or ~0.85% at the total-fund level. Bond selection is unknown.

Cash drag: 4% cash at ~0% versus 7.92% policy costs about 0.32%. Fees of 0.75% reduce gross return. The residual (roughly 0.2-0.4%) is likely fees, cash, and interaction effects, plus the difference between the 8.1% benchmark and the 7.92% policy benchmark.

The client should require a policy benchmark that reflects the strategic allocation in the IPS (60/40), with a separate custom benchmark for any tactical range. The 8.1% benchmark is unexplained and may embed a different allocation or a peer-group average, which makes attribution meaningless. Require the manager to report: (1) total return gross and net of fees; (2) allocation effect, selection effect, and interaction effect using a Brinson-style model with the policy benchmark; (3) a cash and fee reconciliation; (4) a currency effect if global equity is unhedged.

Methodology: Brinson-Fachler or Brinson-Hood-Beebower, with the choice documented. Require the manager to state whether the attribution is arithmetic or geometric and to reconcile the sum of effects to the actual excess return within a stated tolerance (e.g., 5 basis points). Require the manager to disclose the average weights used (time-weighted versus dollar-weighted) and to explain the treatment of cash. Require a GIPS-compliant presentation if the manager claims compliance.

The client should also require a peer-universe comparison only as context, not as the primary benchmark, and should require the manager to explain any quarter where the residual exceeds tolerance.

Sharpe ratio measures total return per unit of total volatility; information ratio measures active return per unit of tracking error. A higher Sharpe with a low information ratio means the fund has delivered good absolute risk-adjusted returns but has not done so efficiently relative to its benchmark. Possible explanations: the fund's beta or allocation drift (65% equity versus 60%) raised both return and volatility in a rising market, so the Sharpe improvement is compensation for beta, not skill; the fund's active bets were large and inconsistent, producing high tracking error with modest active return; or the fund's cash and fee drag reduced active return while its equity tilt reduced measured tracking error.

The advisor should decompose the Sharpe ratio into a beta component and an alpha component, compute the fund's beta versus policy, and compare the information ratio to peers with similar tracking error budgets. If the fund's active risk budget is 4% and its information ratio is 0.2, that is weak; if the budget is 2% and the IR is 0.4, it is more defensible. The advisor should also check whether the benchmark's Sharpe is computed on the same period and frequency, and whether the fund's returns are net of fees. Recommend the client evaluate the manager on information ratio and alpha within the stated risk budget, not on Sharpe alone.