Why Raw Returns Don’t Tell You What’s Working
You beat your benchmark by 3% last year. Congratulations—but do you know why? Was it because you overweighted technology during a tech rally (allocation effect)? Because you picked the best stocks within each sector (selection effect)? Or because you were more invested during up months and more defensive during down months (timing effect)?
The point is: without attribution, you can’t repeat success or fix failure. A trader who thinks their stock-picking is exceptional—when really they just got lucky with sector exposure—will eventually underperform when sector rotations reverse.
Performance attribution quantifies the sources of return difference between your portfolio and a benchmark. It answers the question: Where exactly did my alpha (or underperformance) come from?
The Brinson Model (The Foundation of Attribution)
The Brinson-Fachler attribution model (1985) remains the industry standard. It decomposes active return into three components:
1. Allocation effect: Did you over/underweight the right sectors?
2. Selection effect: Did you pick better stocks within each sector?
3. Interaction effect: The combined impact of allocation and selection decisions
The formula breakdown:
For each sector:
- Allocation effect = (Portfolio weight – Benchmark weight) × (Benchmark sector return – Benchmark total return)
- Selection effect = Benchmark weight × (Portfolio sector return – Benchmark sector return)
- Interaction effect = (Portfolio weight – Benchmark weight) × (Portfolio sector return – Benchmark sector return)
Total active return = Sum of all sector effects
Sector Allocation Effect (Betting on Industries)
The allocation effect measures whether you correctly positioned for sector performance—regardless of your stock picks within those sectors.
Example calculation:
Your portfolio: 25% Technology (benchmark weight: 20%) Benchmark Technology return: +15% Benchmark total return: +10%
Allocation effect = (25% – 20%) × (15% – 10%) = +0.25%
Interpretation: By overweighting technology (a sector that outperformed the benchmark), you added 25 basis points of return.
When allocation effect is positive:
- You overweighted sectors that beat the benchmark
- You underweighted sectors that lagged the benchmark
When allocation effect is negative:
- You overweighted sectors that underperformed
- You underweighted sectors that outperformed
What matters here: Allocation effect captures your macro views—whether you correctly anticipated which parts of the economy would lead or lag.
Selection Effect (Picking Winners Within Sectors)
The selection effect measures your stock-picking skill within each sector, assuming benchmark weights.
Example calculation:
Benchmark weight in Healthcare: 12% Benchmark Healthcare return: +8% Your Healthcare holdings return: +14%
Selection effect = 12% × (14% – 8%) = +0.72%
Interpretation: Your healthcare stock picks outperformed the healthcare sector by 6 percentage points, contributing 72 basis points to your active return.
When selection effect is positive:
- Your stocks within sectors outperformed the sector average
- You identified superior individual companies
When selection effect is negative:
- Your stock picks lagged sector performance
- You would have been better off buying sector ETFs
Why this matters: Many active investors believe they’re skilled stock pickers when their returns actually come from allocation (sector bets). The selection effect isolates true bottom-up skill.
Interaction Effect (The Multiplier)
The interaction effect captures the combined impact of allocation and selection decisions. It’s positive when you both overweight a sector AND pick stocks that outperform within it.
Example:
Overweight Technology by 5% (portfolio 25% vs benchmark 20%) Your tech stocks return +18% vs sector benchmark +15%
Interaction effect = (25% – 20%) × (18% – 15%) = +0.15%
Interpretation: The combination of overweighting tech AND picking outperforming tech stocks added 15 basis points.
Practical insight: Interaction effects are usually small. If your interaction effect dominates, you may have data issues or very concentrated positions.
A Complete Attribution Example
Your portfolio vs S&P 500 for one quarter:
| Sector | Portfolio Weight | Benchmark Weight | Portfolio Return | Benchmark Sector Return |
|---|---|---|---|---|
| Tech | 30% | 28% | +12% | +10% |
| Healthcare | 15% | 13% | +5% | +7% |
| Financials | 10% | 12% | +8% | +6% |
| Consumer | 20% | 22% | +4% | +3% |
| Other | 25% | 25% | +3% | +4% |
Benchmark total return: +6%
Allocation effects:
- Tech: (30%-28%) × (10%-6%) = +0.08%
- Healthcare: (15%-13%) × (7%-6%) = +0.02%
- Financials: (10%-12%) × (6%-6%) = 0%
- Consumer: (20%-22%) × (3%-6%) = +0.06%
- Other: (25%-25%) × (4%-6%) = 0%
- Total allocation: +0.16%
Selection effects:
- Tech: 28% × (12%-10%) = +0.56%
- Healthcare: 13% × (5%-7%) = -0.26%
- Financials: 12% × (8%-6%) = +0.24%
- Consumer: 22% × (4%-3%) = +0.22%
- Other: 25% × (3%-4%) = -0.25%
- Total selection: +0.51%
Interaction effects: (smaller, calculated similarly) = +0.08%
Total active return: +0.75%
The interpretation:
- Your sector bets (allocation) added 0.16%
- Your stock picks (selection) added 0.51%
- Your stock-picking skill was 3x more important than your sector allocation this quarter
Time-Period Attribution (Understanding Timing)
Beyond single-period analysis, examine how attribution changes over time:
Rolling attribution: Calculate monthly attribution and observe trends
- Are your selection effects consistently positive, or randomly positive and negative?
- Does your allocation effect correlate with market regime changes?
Up-market vs down-market attribution:
- Some managers add value only in up markets (high beta)
- Others add value in down markets (defensive selection)
- Understanding your pattern helps set expectations
Contribution timing: Did you add positions before or after the performance occurred?
- Buying a stock that already rallied 30% isn’t skill
- Adding before the move suggests genuine insight
Common Attribution Mistakes
Mistake 1: Survivorship bias in selection effect If you sold your losers and kept your winners, your current portfolio looks more skilled than your actual trading decisions were. Attribution should include all positions held during the period, not just current holdings.
Mistake 2: Ignoring cash drag Cash holdings (even 5%) create negative allocation effect in rising markets. If your benchmark is 100% invested and you hold 10% cash, you’ll underperform in most years by roughly 10% × (equity return – cash return).
Mistake 3: Benchmark mismatch Comparing a small-cap growth portfolio to the S&P 500 creates meaningless attribution. Your benchmark should match your investable universe.
Mistake 4: Single-period conclusions One quarter of strong selection effect doesn’t prove skill. You need 12-24 months of data to distinguish skill from variance.
Building an Attribution System
Manual approach (spreadsheet):
- Export monthly portfolio weights and returns by sector
- Source benchmark sector weights and returns (free from index providers)
- Build formulas for allocation, selection, and interaction effects
- Sum across sectors for total active return decomposition
Automated approaches:
- Portfolio management software (Portfolio Visualizer, Sharesight)
- Broker-provided analytics (Fidelity, Schwab have basic attribution)
- Professional tools (FactSet, Bloomberg for institutional investors)
Minimum viable cadence:
- Quarterly attribution analysis (align with earnings seasons)
- Annual summary with rolling 12-month trends
- Event-driven deep-dives after significant underperformance
Detection Signals (When Attribution Reveals Problems)
Your attribution analysis is signaling issues if:
- Selection effect is consistently negative while you believe you’re a stock picker (you’re not—switch to sector ETFs)
- Allocation effect swings wildly between large positive and negative (your macro views add noise, not value)
- Interaction effect is large relative to allocation and selection (check data quality or position concentration)
- Your best quarters come from allocation, worst from selection (you’re making sector bets, not stock picks)
Mitigation Checklist
Essential (high ROI)
These 4 items give you foundational attribution insight:
- Choose an appropriate benchmark matching your investment universe
- Calculate quarterly attribution decomposing allocation vs selection effects
- Track selection effect over 12+ months to measure true stock-picking skill
- Include all positions held during the period (not just survivors)
High-Impact (systematic approach)
For investors serious about understanding their edge:
- Analyze up-market vs down-market attribution separately
- Compare attribution across time periods to identify consistency
- Attribute cash drag separately from security selection
Optional (for advanced practitioners)
If managing significant capital or professional mandates:
- Use multi-factor attribution (style, country, currency effects)
- Calculate risk-adjusted attribution (did selection effect come with higher risk?)
- Implement transaction-based attribution capturing intra-period timing
Next Step (put this into practice)
Run a basic attribution analysis on your portfolio for the past year.
How to do it:
- List your average weight in each sector for the period
- Source S&P 500 sector weights and returns (free at sectorspdr.com)
- Calculate allocation effect: (your weight – benchmark weight) × (sector return – benchmark return)
- Calculate selection effect: benchmark weight × (your return in sector – sector benchmark return)
- Sum across sectors
Interpretation:
- Positive allocation, positive selection: Skill in both areas
- Positive allocation, negative selection: Sector views are working, stock picks aren’t
- Negative allocation, positive selection: Good stock picker, poor sector allocator
- Negative allocation, negative selection: Both areas need work
Action: If selection effect is consistently negative over 4+ quarters, consider replacing individual stock holdings with sector ETFs in your weakest selection areas. Concentrate your stock-picking in sectors where you demonstrate positive selection effect.
