By Rob Bush, DWS
Great athletes are known for constant self-evaluation and relentless performance analysis. Whether they played well, or poorly, many want as much detail as possible about what worked, and what didn’t, and why. Similarly in finance, much can be learned by analyzing approaches that go off benchmark to figure out what decisions paid off, either as a way to fine tune future performance, or as a vehicle to better understand past performance.
To be sure, attribution can be a tricky business, and lots of methodologies have been created to analyze portfolio performance. In this blog, I plan to briefly discuss two popular methods and provide some thoughts around what they are telling us, and how useful that information might be.
The Attribution Table
Figure One shows an attribution table with some made up numbers for a hypothetical portfolio that differs from the market in its stock and sector picks. Note a couple of obvious features. First, the portfolio and benchmark weights in each sector have to sum to 100% (in reality there could be a cash portion or other buckets but the intent here is to keep things simple). Second, the active weights (the differences in sector weights) have to sum to 0% – for every overweight there has to be an underweight.
Figure One: A hypothetical attribution table with example numbers
Next, note that the portfolio return of 3.59% is simply the sum product of the portfolio weights and returns, and the same is true for the benchmark return of 3.29%. This manager has managed to outperform their index by 0.30% over the period. The question is – where did those 30 basis points come from? What were the sector and stock level positions that won and lost? That’s ultimately what attribution is trying to answer.
Contribution to Return
The two sections that carve up the 30 basis points are labelled Contribution and Attribution. The former is very straightforward and, in our view, serves quite well as an indicator of what was going on (note in passing it’s pretty similar to the final column which is the sum of the more detailed attribution, more on that later).
The portfolio contribution to return is simply the weight of the sector in the portfolio multiplied by the return of the sector in the portfolio (note this will differ to the return to the benchmark because of the different stock holdings). And, similarly, the benchmark return is the weight of the sector in the benchmark multiplied by the return of the sector in the benchmark. Difference in contribution to return (here the column labelled “Active” in the Contribution section) is obviously just one minus the other (portfolio minus benchmark). Finally, the sum of both the difference in contribution to return, and the sum of the more detailed attribution are both 30 basis points, as one would expect.
That it’s a fairly intuitive measure is clear when you look at the highest and lowest results in the Active column. Industrials had the best active contribution to return score because the PM was overweight in a sector that went up, and managed to pick superior stocks (portfolio return was 2% higher than the benchmark return). Energy on the other hand had the most negative contribution to return difference because, although the PM was underweight a sector that did poorly, their stock picks within the sector were bad.
However, note an example of where we think focusing solely on difference in contribution to return can mask interesting information. For the Health Care sector the portfolio, and the benchmark, contributions to return were both 0.50% (and hence the difference in contribution to return was 0.00%). That would suggest, by looking at this metric alone, that there was nothing of interest going on there. But, isn’t it apparent from a glance at the PM’s strongly overweight position in the Health Care sector, and their disappointing stock picking performance, that that wasn’t the case. The sector was a high conviction trade (highest active overweight), but the stock picking performance was bad (second worst, after Energy). So it’s probably fair to say that, while it may be a reasonable first cut of the data, contribution to return can run the risk of overlooking meaningful information.
In part two of this blog, I’ll discuss Brinson attribution, a more granular approach to portfolio analysis that solves for this type of problem by getting under the hood of decisions at three different levels. Until then, take your lead from those high performing athletes, and keep evaluating your performance.