Learning Outcome Statement
Describe returns-based, holdings-based, and transactions-based performance attribution, including advantages and disadvantages of each.
On the exam: You get a manager-and-benchmark case or method descriptions and must compare the three approaches, including advantages and disadvantages. The response may also require a short holdings-based active-return reconciliation.
Your first decision is not a formula. It is a data decision. The record available to you controls how directly the result can be tied to a manager action.
| Approach | Data used | Trade-off |
|---|---|---|
| Returns-based | Portfolio and benchmark returns | Fast and inexpensive, but exposures are inferred |
| Holdings-based | Portfolio and benchmark holdings | Transparent, but timing can be missed |
| Transactions-based | Holdings, trades, and cash flows | Timing captured, but complete, accurate records required |
Returns-based methods can work when holdings are unavailable, but return patterns can fit several exposures. Holdings-based methods show what the portfolio and benchmark held at measurement dates, while transactions-based methods improve treatment of trading timing and cash flows but require complete, accurate transaction records.
The arithmetic reconciliation below is the holdings-based approach. It uses beginning-of-period portfolio and benchmark weights, group returns, and a gross-of-fees convention.
For the reconciliation, use this notation.
| Symbol | Definition |
|---|---|
| $i$ | Group index |
| $w_{p,i}$ | Portfolio weight in group $i$ |
| $w_{b,i}$ | Benchmark weight in group $i$ |
| $r_{p,i}$ | Portfolio return for group $i$ |
| $r_{b,i}$ | Benchmark return for group $i$ |
| $R_p$ | Total portfolio return |
| $R_b$ | Total benchmark return |
The difference $w_{p,i}-w_{b,i}$ is the active weight. The difference $r_{p,i}-r_{b,i}$ is the group return spread. The formula below decomposes active return into allocation, selection, and interaction effects.
$$ \text{Active return} = \sum_i \left[(w_{p,i}-w_{b,i})(r_{b,i}-R_b) + w_{b,i}(r_{p,i}-r_{b,i}) + (w_{p,i}-w_{b,i})(r_{p,i}-r_{b,i})\right] $$
Read the products from left to right as allocation, selection, and interaction. Allocation is positive when the active weight and benchmark relative return share a sign, while selection uses the benchmark weight to scale the group return spread. Interaction is the joint product of active weight and that spread. The three effects add to $R_p-R_b$ under this one-period arithmetic convention.
| Measure | Inputs | Output |
|---|---|---|
| Contribution | Portfolio weight and portfolio return | Portfolio's own return |
| Attribution | Portfolio and benchmark returns | Active return |
A contribution can sum to the portfolio's own return, but it is not the benchmark-relative reconciliation. For the benchmark-relative question, trace each effect to the pair of quantities that creates it. Allocation asks whether the portfolio held more or less than the benchmark in groups that beat or lagged the benchmark overall. Selection asks whether each group beat its own benchmark group, scaled by the benchmark's weight. Interaction captures the combined consequence of changing both weight and return. Keeping those drivers distinct helps explain a result instead of merely reporting its total.
| Effect | Formula term | Driver pair |
|---|---|---|
| Allocation | $(w_{p,i}-w_{b,i})(r_{b,i}-R_b)$ | Active weight, benchmark relative return |
| Selection | $w_{b,i}(r_{p,i}-r_{b,i})$ | Benchmark weight, group return spread |
| Interaction | $(w_{p,i}-w_{b,i})(r_{p,i}-r_{b,i})$ | Active weight, group return spread |
Under this arithmetic convention, add the effect totals to the benchmark return to reach the portfolio return. This bridge is useful only when the inputs and convention are stated consistently. A return pattern can suggest exposures, but it cannot show precisely what the portfolio held on a date. Holdings provide that snapshot, though they may miss trading that happened between measurement dates. Transaction records add that timing detail and the associated cash flows. The extra detail improves the account of manager activity only when the underlying records are complete and accurate, so match the method to evidence the analyst can defend.
Choose the richest path the records support, not a more detailed path than the records can defend. The image summarizes how the three methods rely on increasingly detailed records, while the decision path below turns that comparison into a practical choice. Start with what is actually available: returns alone support an inferred exposure analysis; holdings permit a more direct comparison at recorded dates; and trades with cash flows allow trading timing to enter the analysis. More data does not remove the need to check its quality. If transaction records are incomplete, the apparent precision of a transactions-based result can be misleading. State the limitation along with the result.
Loading diagram…
Worked example
You are handed a one-period equity portfolio and benchmark split between groups A and B. The investment committee wants to know whether the portfolio's active return came from allocation, selection, or interaction. Use beginning-of-period weights, one-period arithmetic attribution, and gross-of-fees returns.
| Input | Value |
|---|---|
| Benchmark weights | $w_{b,A}=0.5$, $w_{b,B}=0.5$ |
| Portfolio weights | $w_{p,A}=0.6$, $w_{p,B}=0.4$ |
| Benchmark returns | $r_{b,A}=0.1$, $r_{b,B}=0.06$ |
| Portfolio returns | $r_{p,A}=0.12$, $r_{p,B}=0.07$ |
The bars are cumulative stages, so the bridge starts at the benchmark return and ends at the portfolio return. Before reading the chart, keep the distinction between each stage's incremental effect and its cumulative level in view. The benchmark is the starting point, allocation moves that base by 0.0040, selection adds another 0.0150, and interaction contributes the final 0.0010. The last bar therefore shows the portfolio return, while the three changes together explain the active return. This ordering makes it easier to audit the arithmetic and identify which driver matters most in this example.
Loading diagram…
| Stage | Change | Cumulative return |
|---|---|---|
| Benchmark | Base | 0.0800 |
| Allocation | 0.0040 | 0.0840 |
| Selection | 0.0150 | 0.0990 |
| Interaction | 0.0010 | 0.1000 |
- Build the benchmark total first because allocation uses it as the reference. The benchmark return is $R_b=0.5 \times 0.1+0.5 \times 0.06=0.0800$.
- Build the portfolio total and active return so the effects have a reconciliation target. The portfolio return is $R_p=0.6 \times 0.12+0.4 \times 0.07=0.1000$, and active return is $R_p-R_b=0.1000-0.0800=0.0200$.
- Calculate the active weights and return spreads because each effect is a product of these drivers. The outputs are $w_{p,A}-w_{b,A}=0.1000$, $w_{p,B}-w_{b,B}=-0.1000$, $r_{b,A}-R_b=0.0200$, $r_{b,B}-R_b=-0.0200$, $r_{p,A}-r_{b,A}=0.0200$, and $r_{p,B}-r_{b,B}=0.0100$.
- Multiply the paired drivers and keep the effects separate for the committee. Allocation is $0.1000 \times 0.0200+(-0.1000) \times (-0.0200)=0.0040$. Selection is $0.5 \times 0.0200+0.5 \times 0.0100=0.0150$. Interaction is $0.1000 \times 0.0200+(-0.1000) \times 0.0100=0.0010$.
- Reconcile the effects as an audit check. The total is $0.0040+0.0150+0.0010=0.0200$, matching $R_p-R_b$.
The committee should record portfolio return $0.1000$ against benchmark return $0.0800$, giving active return $0.0200$ led by selection at $0.0150$, while treating the outcome as evidence about this period rather than proof of persistent manager skill.
Changed condition
Keep every weight and benchmark return unchanged. Group A's portfolio return falls from 0.12 to 0.1. Predict the new active-return reconciliation before recalculating.
| Measure | Before | After |
|---|---|---|
| Group A portfolio return | 0.12 | 0.1 |
| Selection effect | 0.0150 | 0.0050 |
| Interaction effect | 0.0010 | -0.0010 |
Your prediction should be that allocation stays at 0.0040 because weights and benchmark returns did not move. Selection falls from 0.0150 to 0.0050 because group A's return now matches its benchmark return. Interaction moves from 0.0010 to -0.0010 because group A's spread disappears while group B's interaction remains -0.0010. The portfolio return becomes 0.0880, so active return falls from 0.0200 to 0.0080. Before reading on, recompute the changed interaction effect from the group-level inputs yourself.
Common error
You may see a sector's return contribution in your portfolio and call it an attribution effect. That feels right because the sector added to your portfolio's own return. The distinction is that contribution uses the portfolio weight and portfolio return, while attribution analysis explains your active return relative to the benchmark and must reconcile under the stated convention.
Go deeper
Arithmetic attribution adds the effects within the stated period. A geometric attribution links periods through compounding, so its effects should not be inserted into this additive reconciliation. Keep the gross or net label consistent across the portfolio return, benchmark return, and effects.
Check yourself
Practice
describe
Describe the best attribution approach when holdings, trades, and cash flows are complete, and identify its main disadvantage.
Which statement best distinguishes contribution from attribution?
Revision snapshot
The holdings-based arithmetic formula is below.
$$ R_p-R_b=\sum_i\left[(w_{p,i}-w_{b,i})(r_{b,i}-R_b)+w_{b,i}(r_{p,i}-r_{b,i})+(w_{p,i}-w_{b,i})(r_{p,i}-r_{b,i})\right] $$
Returns-based attribution infers exposures from return patterns. Holdings-based attribution uses portfolio and benchmark holdings. Transactions-based attribution uses holdings plus trades and cash flows. Holdings-based attribution commonly decomposes active return into allocation, selection, and interaction effects under a stated arithmetic convention. Attribution explains active return relative to a benchmark. Contribution is the portfolio's own return contribution and is not the same measure. Transactions-based attribution improves treatment of trading timing and cash flows but requires complete, accurate transaction records. On the exam, name the data used and its trade-off, state the convention, and check the reconciliation. One strong attribution period is evidence about that period, not proof of persistent manager skill.
Unlock the full lesson and continue reading
Signup to continue reading this lesson and unlimited access questions, flashcards, AI notes, and more