Performance Attribution: Where a Portfolio's Return Actually Came From

Beating a benchmark by two percentage points tells you nothing on its own about whether a manager is skilled at picking sectors, skilled at picking stocks within a sector, or simply got lucky with one oversized bet. Performance attribution decomposes total excess return into its distinct sources, and that decomposition is what tells you whether an active fee is buying something repeatable.

By Karanvir Singh, founder of Quantix Invest

The core idea: allocation effect versus selection effect

The standard Brinson-style attribution framework splits a portfolio's excess return, its total return minus the benchmark's total return, into two primary components. The allocation effect measures the value added or subtracted purely by weighting sectors or asset classes differently than the benchmark does, independent of which specific securities were chosen within each sector. The selection effect measures the value added or subtracted by choosing better or worse individual securities within each sector, holding the sector weight itself constant at the benchmark's weight.

The mechanism that makes this decomposition useful is that the two effects require entirely different skills and carry entirely different implications for what an investor should pay for. Allocation calls, overweighting technology because you expect it to outperform, are broad macro or sector bets that can usually be replicated cheaply and directly with low-cost sector ETFs. Selection calls, picking the specific technology companies that will outperform the technology sector average, require deeper research and are harder to replicate passively, which is the kind of skill an active management fee is more plausibly compensating.

The math: decomposing a two-sector portfolio

Consider a benchmark composed of 30% technology, returning 15%, and 70% utilities, returning 3%. The benchmark's total return is (0.30 × 15%) + (0.70 × 3%) = 4.5% + 2.1% = 6.6%. A manager instead holds 50% technology, where their specific technology picks return 14%, and 50% utilities, where their specific utility picks return 4%. The manager's total portfolio return is (0.50 × 14%) + (0.50 × 4%) = 7% + 2% = 9%, beating the benchmark by 9% − 6.6% = 2.4%.

The allocation effect for each sector is the sector's weight difference multiplied by the sector's own benchmark return minus the overall benchmark return: for technology, (0.50 − 0.30) × (15% − 6.6%) = 0.20 × 8.4% = 1.68%. For utilities, (0.50 − 0.70) × (3% − 6.6%) = (−0.20) × (−3.6%) = 0.72%. The total allocation effect is 1.68% + 0.72% = 2.40%. The selection effect for each sector is the benchmark's weight in that sector multiplied by the difference between the manager's return in that sector and the benchmark's return in that sector: for technology, 0.30 × (14% − 15%) = 0.30 × (−1%) = −0.30%. For utilities, 0.70 × (4% − 3%) = 0.70 × 1% = 0.70%. Total selection effect is −0.30% + 0.70% = 0.40%.

Here the allocation effect of 2.40% and the selection effect of 0.40% sum to 2.80%, slightly above the actual 2.4% excess return, because the standard Brinson framework also has a third term, the interaction effect, capturing the combined impact of overweighting a sector and picking better or worse securities within it simultaneously. For technology, interaction is (0.50 − 0.30) × (14% − 15%) = 0.20 × (−1%) = −0.20%; for utilities, (0.50 − 0.70) × (4% − 3%) = (−0.20) × 1% = −0.20%, a total interaction effect of −0.40%. Now the three components sum correctly: 2.40% + 0.40% + (−0.40%) = 2.40%, matching the actual excess return exactly. The full picture shows the manager's outperformance came almost entirely from the sector bet on technology, not from stock-picking, and that a modest, offsetting stock-picking weakness within technology itself was more than made up for by strong picks within the smaller utilities sleeve.

A second example: a three-sector portfolio and an interaction term

Extend the framework to three sectors to see how attribution scales. A benchmark holds 40% technology (returning 10%), 35% healthcare (returning 6%), and 25% energy (returning −4%). Benchmark total return: (0.40 × 10%) + (0.35 × 6%) + (0.25 × −4%) = 4% + 2.1% − 1% = 5.1%. A manager holds 25% technology (returning 12%), 50% healthcare (returning 7%), and 25% energy (returning −2%). Manager's total return: (0.25 × 12%) + (0.50 × 7%) + (0.25 × −2%) = 3% + 3.5% − 0.5% = 6%, an excess return of 6% − 5.1% = 0.9%.

Allocation effects: technology, (0.25 − 0.40) × (10% − 5.1%) = (−0.15) × 4.9% = −0.735%; healthcare, (0.50 − 0.35) × (6% − 5.1%) = 0.15 × 0.9% = 0.135%; energy, (0.25 − 0.25) × (−4% − 5.1%) = 0, since the manager held the same weight as the benchmark. Total allocation effect: −0.735% + 0.135% + 0% = −0.60%, meaning the manager's sector weighting actually subtracted value, mostly from being underweight the strong technology sector.

Selection effects, using benchmark weights: technology, 0.40 × (12% − 10%) = 0.40 × 2% = 0.80%; healthcare, 0.35 × (7% − 6%) = 0.35 × 1% = 0.35%; energy, 0.25 × (−2% − −4%) = 0.25 × 2% = 0.50%. Total selection effect: 0.80% + 0.35% + 0.50% = 1.65%, strongly positive across all three sectors. This example shows the opposite pattern from the first: here the manager's sector timing actively hurt performance, being underweight the sector that ran hardest, but genuinely superior stock selection within every single sector more than compensated, producing a positive total excess return built on exactly the kind of skill, security selection, that is harder to replicate and more plausibly worth an active fee.

What the evidence shows about where returns come from

Studies applying attribution analysis across large samples of actively managed equity funds have generally found that a substantial share of measured excess return, in many samples the majority, traces back to allocation and timing decisions rather than security selection, and that the selection component specifically has tended to be smaller, noisier, and less consistent across managers and time periods than the allocation component. This matters directly for fee evaluation, since allocation tilts are the piece of the return an investor could largely replicate on their own with a handful of low-cost sector or factor ETFs, at a fraction of an active management fee.

A separate and important finding across long-run attribution studies is that selection effects, when they are genuinely positive, show more persistence from one period to the next than allocation or timing effects do, consistent with the broader finding that stock-picking skill, where it exists at all, tends to be a more stable, repeatable characteristic of a manager than sector-timing skill. This is the empirical basis for treating a consistently positive selection effect across multiple, separate measurement periods as a more credible signal of genuine, ongoing skill than a single strong allocation-driven year, however impressive that single year's headline number looks.

It is worth pausing on why the sign of the interaction term flipped between the two worked examples, since it illustrates a subtlety attribution reports do not always explain clearly. In the first example, the manager overweighted the sector that also happened to be where their stock-picking was weakest, technology, producing a negative interaction effect that partially offset the strong allocation call. In the second example, the manager underweighted the sector where their stock-picking was actually strongest relative to the benchmark, technology again, which also produced a negative interaction effect, but this time compounding an already-negative allocation call rather than offsetting a positive one. The interaction term is neither good nor bad on its own; it simply captures how the sizing decision and the selection decision reinforced or worked against each other within the same sector, and a manager reviewing their own attribution history for patterns should look specifically at whether their interaction term is consistently negative, which can indicate a subtle tendency to overweight sectors right before their own picks within that sector underperform, a correctable behavioral pattern once identified.

Using attribution to evaluate a manager

For an investor deciding whether to keep paying an active management fee, the practical workflow is to request or reconstruct a multi-year attribution history, not just a single-period snapshot, and specifically track the selection effect's sign and consistency separately from the allocation effect's sign and consistency. A manager whose entire multi-year track record of outperformance comes from allocation, correctly overweighting sectors that happened to run, but whose selection effect is flat or negative most years, is offering a service replicable with sector ETFs at a much lower cost, no matter how good the headline total return numbers look on a fact sheet.

This distinction matters most for high earners choosing among actively managed funds inside taxable brokerage accounts or self-directed retirement accounts, where the fee difference between an active fund and a comparable index-based alternative compounds meaningfully over a multi-decade horizon. A fund charging a materially higher expense ratio needs to demonstrate a persistent, positive selection effect, not just a positive total excess return, to justify that ongoing cost difference over holding a low-cost index fund plus, if desired, a handful of low-cost sector tilts assembled directly by the investor.

Actionable breakdown

• Ask an active manager for a full attribution report, not just total return.
  • Separate allocation, selection, and interaction effects explicitly.
  • Confirm the three components sum to the actual excess return.
• Weigh selection effect more heavily than allocation effect.
  • Selection skill has shown more persistence across periods historically.
  • Allocation tilts are usually replicable cheaply with sector ETFs.
• Check consistency across multiple, separate measurement periods.
  • One strong year is weak evidence of a repeatable process.
  • Several consecutive years of positive selection is a stronger signal.
• Compare the fee to the specific value of the selection effect alone.
  • Do not pay an active fee for a return you could replicate passively.
  • Judge the fee against selection effect, not total excess return.

Common pitfalls

The most common pitfall is crediting a manager's overall skill for outperformance that was really one lucky, concentrated sector bet, unlikely to repeat in the same direction next period. A second pitfall is ignoring that allocation effects can usually be replicated cheaply with low-cost sector ETFs, meaning an investor may be paying an active fee for something they could largely assemble themselves at a fraction of the cost.

A third pitfall is judging a single period's attribution in isolation instead of looking across several separate periods, since one strong quarter or year of any single effect, allocation or selection, can easily be statistical noise rather than a repeatable pattern. A fourth pitfall is ignoring the interaction term entirely and treating allocation plus selection as a complete explanation, which can misstate the true source of return whenever a manager both shifted sector weights and changed security selection quality within those sectors at the same time.

The bottom line

Before paying for active management, confirm the outperformance traces to a persistent, positive selection effect across multiple periods, not a one-time allocation bet you could have replicated yourself.

Related reading: evaluating stocks and managers, factor investing, measuring performance when composition changes, style analysis for fund evaluation, why market timing rarely works.

All articles · The deep guides