Security selection is the choice of securities, and their relative holdings, within an investment category. Choosing which shares to hold inside an equity allocation or which bonds to hold inside a bond allocation can change performance even when the overall mix of asset classes stays fixed. For a quant portfolio, a model's security rankings matter only through the positions they produce and the returns those positions earn.
Asset allocation sets how much capital goes into categories such as equities and bonds. Selection decides what is held inside them. These decisions can be made by people, systematic rules or a combination. A manager can also choose different weights among securities that appear in the benchmark portfolio; selection does not require discovering a company outside the index.
The boundary depends on the classification used. Moving capital between equity sectors is sector allocation in a sector-based report. Choosing stocks inside a sector is selection at that level. Specify the categories before interpreting the result.
Hold the allocation fixed
Consider a hypothetical, fully invested portfolio and benchmark over one common period. Both begin with 60% in equities and 40% in bonds. There are no trades or external cash flows during the period. All returns are gross total returns, in the same currency, with fees and taxes ignored.
| Asset class | Portfolio and benchmark beginning weight | Benchmark class return | Portfolio class return |
|---|---|---|---|
| Equities | 60% | 10% | 12% |
| Bonds | 40% | 2% | 1% |
The equity holdings beat their class benchmark; the bond holdings fall behind theirs. Beginning portfolio weights give the overall returns:
Active return, defined here as the arithmetic difference, is +0.8 percentage points, or 80 basis points. That advantage did not come from allocating more to equities: both portfolios started with the same class weights.
Original synthetic one-period example: beginning weights, gross total returns in one currency, no trading, external flows, fees or taxes. Contributions use the stated three-effect Brinson–Fachler convention. The categories are asset classes; the same calculation can use an agreed sector classification.
Name the attribution convention
For this example, use one-period arithmetic Brinson–Fachler attribution with three separate effects. Let and be portfolio and benchmark beginning weights, and their category returns, and . Both sets of weights sum to one. Define
Here is allocation, is selection and is interaction. Together they reconcile to . These are the three-effect formulas documented in Carl Bacon's attribution review, p. 12.
Because in each category, both allocation and interaction are zero. Selection is
The equity choice adds 1.2 percentage points to total portfolio return relative to the benchmark, while the bond choice subtracts 0.4 points. The net is +0.8 points. The equity sleeve's 2 percentage point advantage is diluted by its 60% weight; it is not a 2-point gain for the whole portfolio.
Some reports combine interaction with selection. When category weights differ, that convention can change the reported selection number. Compare definitions before comparing managers. With equal category weights, interaction is zero under the convention above, so that distinction disappears in this example.
A contribution is not proof of skill
The result locates the realised advantage inside the asset classes. It does not establish that the choices were repeatably skilful or produced risk-adjusted alpha. For example, choosing longer-duration bonds can change interest-rate exposure even though the overall bond allocation is unchanged. Equity selection can change factor exposure, concentration and liquidity risk.
A useful review asks whether the class benchmarks fit the holdings, whether added risks explain the result, whether the advantage survives costs, and whether it persists across an adequately examined record. CFA Institute's performance evaluation summary distinguishes describing return contributions from appraising the investment process.
For the next analytical step, Risk-Adjusted Performance Measures develops Jensen's alpha and multi-index comparisons (membership for the full lesson; public preview available). Those models ask a different question from the fixed-category attribution above: how much return remains after accounting for specified systematic exposures?
Sources
- Carl Bacon, Performance Attribution: History and Progress, CFA Institute Research Foundation, 2019, pp. 12–15 — the three-effect Brinson–Fachler convention and the alternative of combining interaction with selection.
- CFA Institute, Portfolio Performance Evaluation (2026), public introduction and summary — attribution, appraisal, benchmark suitability and risk.
- Gary P. Brinson, L. Randolph Hood and Gilbert L. Beebower, “Determinants of Portfolio Performance,” Financial Analysts Journal 42(4), 1986, pp. 39–41 — policy weights versus active selection within asset classes. Local text checked; the numerical example here is synthetic.