Case 069Manager evaluation and attributionHard

The Rivanta pod beat its benchmark by 2%. It overweighted banks (30% against 20%) and underweighted IT (10% against 25%), with its own stock returns in each sector given. Split the excess return into allocation and selection effects.

The situation

The Rivanta pod runs a long-only sleeve against an Indian large-cap benchmark on a multi-manager platform. Last year it returned 11.75% against the benchmark's 9.75%, 2.00 points ahead.

Weights: banks 30% in the pod against 20% in the benchmark; IT 10% against 25%; all other sectors 60% against 55%. Returns: the pod's banks returned 12% against 15% for the benchmark's banks; its IT stocks 8% against 5%; its other stocks 12.25% against 10%.

Your task

Split the 2.00 points into allocation, choosing how much to hold in each sector, and selection, choosing which stocks to hold within it, by sector. What does the split say about the PM's skill?

Quick check

The pod was overweight banks and its banks lagged the benchmark's banks. What was the net effect of the banks decision?

Worked solution

Try it on paper, then open one step at a time.

30-second answerThe answer to give first

Allocation added 1.25 points and selection 0.75, making the 2.00. Overweighting banks and underweighting IT were both right, adding 0.52 and 0.71. Stock picking was mixed: the pod's banks lagged and cost 0.90, while its other holdings added 1.35. The PM's sector calls worked; the bank picks did not, and a good year hides that.

Step 1What two decisions does attribution separate?

How much to put in each sector, and what to buy inside it. A family choosing how to split a grocery budget between vegetables and snacks makes one decision; choosing which vegetables at which stall is another, and either can be right while the other is wrong. The Brinson attributionA method, from Brinson, Hood and Beebower and later Brinson and Fachler, that splits a portfolio's excess return into the effect of sector weights and the effect of stock choices within sectors. measures allocation as the extra weight in a sector times how much that sector beat the benchmark as a whole, and selection as the pod's weight times how much its stocks beat the benchmark's stocks in the same sector. The benchmark as a whole returned 9.75%, which is the hurdle every sector is measured against.

What the pod held and what its holdings returned, against the benchmarkWeight in the book, %BanksPod 30Benchmark 20ITPod 10Benchmark 25OthersPod 60Benchmark 55Weights sum to 100 on both sidesReturn of the sector's stocks, %BanksPod 12Benchmark 15ITPod 8Benchmark 5OthersPod 12.25Benchmark 10Wholebenchmark9.75: the hurdle
Rivanta held 30% in banks against 20%, 10% in IT against 25% and 60% elsewhere against 55%, and its stocks returned 12%, 8% and 12.25% against the benchmark's 15%, 5% and 10% in the same sectors.
Step 2What are the numbers, sector by sector?

Work each sector twice. Banks: the extra 10 points of weight times banks' 5.25-point lead over the benchmark gives +0.525 of allocation; 30% times the pod's 3-point shortfall inside banks gives -0.90 of selection. IT: 15 points underweight a sector that trailed by 4.75 points gives +0.7125; 10% times a 3-point lead gives +0.30. Others: 5 points over a sector barely ahead of the benchmark adds only +0.0125; 60% times a 2.25-point lead adds +1.35.

SectorAllocationSelectionTotal
Banks+0.525-0.900-0.375
IT+0.712+0.300+1.012
Others+0.012+1.350+1.362
Total+1.250+0.750+2.000
Rivanta's 2.00 points split into 1.25 of allocation, almost all from the banks overweight and IT underweight, and 0.75 of selection, where strong picks outside banks and IT outweighed a -0.90 drag from the bank picks.
The 2.00-point excess, split by decision and sector+0.52Banksallocation+0.71ITallocation+0.01Othersallocation-0.90Banksselection+0.30ITselection+1.35Othersselection2.00ExcessreturnAllocation +1.25Selection +0.75Percentage points of return against the benchmark
Rivanta's 2.00-point excess return comes from 1.25 points of allocation and 0.75 of selection; in banks the overweight added 0.52 while the choice of banks took away 0.90, so the two effects pulled in opposite directions.
Step 3What does the split say about the PM, and what are its limits?

It tells you where to probe. The sector calls were right and worth 1.25 points; the bank picks were wrong and cost 0.90. A platform might ask whether the bank picks had a common tilt, such as smaller lenders, that lost to the large banks. Two limits deserve a sentence. First, this version folds the interaction effect into selection; measuring selection on benchmark weights instead gives 1.39 of selection and -0.64 of interaction, the same total split differently, so say which convention you used. Second, one year of attribution describes what happened, not whether it was skill.

Where candidates lose it

The frequent error is measuring allocation against the sector's absolute return instead of its return relative to the whole benchmark. Overweighting a sector that rose 10% is not a good call if the benchmark rose 12%.

The second is netting everything into one number per sector. The point of the exercise is that allocation and selection in banks pulled in opposite directions; a single -0.38 for banks hides both.

What the interviewer asks next

  • Recompute with selection on benchmark weights and interaction shown separately. Which convention would you present to a PM?
  • The pod's bank picks were all small lenders. How would you test whether size, not stock picking, drove the drag?
  • How would attribution change for a long-short pod with short positions?

Company names and figures are illustrative.