CIPM Level 2 Attribution Disagreement: Why Different Models Produce Different Results
- Kateryna Myrko
- 21 hours ago
- 4 min read
Performance attribution is designed to explain why a portfolio performed differently from its benchmark. Yet two analysts can use the same portfolio return and still report different allocation, selection, interaction, or currency effects.
This does not necessarily mean that one analysis is wrong. Attribution results depend on the model’s assumptions, benchmark, calculation method, and treatment of the manager’s investment decisions.
For 2026, the CIPM Level II curriculum requires candidates to understand arithmetic and geometric multiperiod attribution, multicurrency attribution, benchmark effects, and the major approaches to fixed-income attribution.
Attribution Is a Model, Not a Single Universal Answer
A portfolio’s total return is observable. The division of that return into explanatory effects is constructed by an attribution model.
Each model asks a particular question. One may focus on sector-allocation and security-selection decisions. Another may separate currency management. A fixed-income model may divide performance into duration, yield-curve, spread, and security effects.
Different questions naturally produce different explanations.
The appropriate model should reflect the manager’s actual investment process. CFA Institute research emphasizes that attribution results should be aligned with the decisions the manager was responsible for making.
Brinson Models Can Assign Allocation Differently
The Brinson framework commonly separates active performance into:
Allocation
Security selection
Interaction
However, the Brinson–Hood–Beebower and Brinson–Fachler versions define allocation differently.
Under the Brinson–Hood–Beebower approach, an overweight position in a sector with a positive benchmark return can create a positive allocation effect.
Under the Brinson–Fachler approach, the sector’s benchmark return is compared with the overall benchmark return. An overweight position contributes positively only when that sector outperforms the total benchmark. CFA Institute notes that the Brinson–Fachler method is generally better aligned with many investment decision processes.
Therefore, the total active return may remain the same while the allocation effect differs between the two models.
Interaction May Be Reported Separately or Reassigned
Interaction measures the combined effect of allocation and selection decisions.
Suppose a manager both overweights a sector and selects securities within that sector that outperform. Part of the active return results from the two decisions occurring together.
Some attribution reports present interaction as a separate effect. Others combine it with security selection. CFA Institute research explains that combining interaction with selection may be appropriate when the investment process assumes that allocation decisions are made first and security selection follows.
Two reports may therefore show different selection effects even though they reconcile to the same overall active return.
Candidates should look at where the interaction effect has been assigned before concluding that the models disagree.
Benchmark Selection Changes the Entire Explanation
Attribution measures portfolio decisions relative to a benchmark. Changing the benchmark changes:
Active portfolio weights
Active returns
Allocation effects
Selection effects
Tracking error
Information ratios
The 2026 CIPM Level II outline specifically requires candidates to understand the effect of benchmark selection on attribution analysis.
For example, a global equity manager may appear to add value through country allocation when measured against a broad global index. Compared with a strategy benchmark that already reflects the manager’s normal country positioning, the same result may instead be attributed largely to security selection.
Neither result can be interpreted properly without understanding what the benchmark represents.
Arithmetic and Geometric Attribution Do Not Behave the Same Way
Arithmetic attribution expresses effects as additive contributions to arithmetic excess return. It is often intuitive for a single measurement period.
The difficulty appears across multiple periods. Returns compound geometrically, so adding monthly arithmetic attribution effects does not generally equal the difference between the portfolio’s compounded return and the benchmark’s compounded return.
A multiperiod arithmetic analysis therefore requires a linking or smoothing method. Different linking methods may distribute the adjustment across periods and effects differently.
Geometric attribution is designed around relative compounded performance. It may reconcile naturally across periods, but its effects are interpreted differently from arithmetic contributions.
Consequently, an arithmetic report and a geometric report may provide different effect values while both remain internally consistent.
Currency Treatment Can Create Further Differences
In a multicurrency portfolio, performance may reflect:
Local-market asset returns
Currency movements
Currency hedging
Interest-rate differentials
Allocation across countries or currencies
The 2026 curriculum includes the Karnosky–Singer approach and geometric multicurrency attribution. It also requires candidates to understand why interest-rate differentials matter.
A model that treats currency as a separate management decision can produce different allocation and selection effects from a model that embeds currency movements within local asset returns.
Candidates must identify whether the manager controlled currency exposure and how the model separates the relevant decisions.
Fixed-Income Models Use Different Explanatory Structures
Fixed-income attribution can also differ because the models analyse bond performance in different ways.
The official Level II outline identifies three major approaches:
Exposure decomposition using duration
Yield-curve decomposition using duration
Yield-curve decomposition using full repricing
These approaches differ in complexity, required data, assumptions, implementation, and interpretation.
A duration-based model may approximate the effect of interest-rate changes, while full repricing can capture more complex changes in the yield curve and security valuation. The resulting effects may therefore differ even when applied to the same bond portfolio.
Final Exam Takeaway CIPM Level 2 Attribution Disagreement
When two attribution reports disagree, do not immediately search for a calculation error. First determine whether they use the same:
Benchmark
Attribution model
Decision hierarchy
Interaction treatment
Arithmetic or geometric methodology
Multiperiod linking process
Currency framework
Valuation data and classification structure
The total portfolio and benchmark returns may be identical while the explanation of active performance differs. CIPM Level 2 Attribution Disagreement
For CIPM Level II candidates, the central principle is that attribution must match the investment process. The best model is not simply the one producing the most detailed output; it is the one that assigns performance to the decisions the manager was actually responsible for making.
Unlock your potential with our comprehensive CIPM Exam - practice exams and study packages!
Comments