Rebecca,

Indeed Carl is correct in his point that the sector attribution equations can be applied at the stock level, and that the resultant "asset allocation" (from the sector attribution model) actually explains stock selection (in the stock attribution model). The stock selection and interaction terms sum to what I call the "transaction cost" or "intra-day timing" effect in the stock attribution model. Some people simply dispense with this latter effect, and treat it as a residual.

Please see the following two pages:
http://www.compoundinghappens.com/attrib_sector.htm
http://www.compoundinghappens.com/attrib_stock.htm
for my accounts of sector attribution and stock-level attribution respectively.

An additional important point, which Carl didn't cover, is that it's not to hard to calculate a stock-level "drill down" for the stock selection attribute in a sector attribution model. To my mind, it makes sense that the stock selection attribute can be explained down to the stock level, but that asset allocation and interaction can not, since the whole concept of asset allocation is that it takes place at the sector level.

On the stock-level attribution page, you will find an example Attribution2.xls, which shows a sector attribution calculated normally on one sheet, and with a drill-down to stock level in the other sheet. Please have a quick look at that example. I am planning to add a new page about how to calculate sector attribution down to the stock-level as a sub-page of the current sector attribution page. This may take a few weeks to happen. But I think you should get the basic idea from Attribution2.xls.

Please let me know if you have any further comments or questions.

Damien.

__________________
Damien Laker is one of the world's leading authorities in performance analytics. For more information visit Moderators

Damien,

There's an issue in the Exact Method that I don't understand.
I'm doing multiperiod performance attribution at a sector level, and as I understand, the Exact Method uses the BHB version for calculating allocation at a sector level.

I'm using the following basis:

BHB:
AA = (PortfolioSectorWeight - BenchmarkSectorWeight) * BencharkSectorReturn

BF:
AA = (PortfolioSectorWeight - BenchmarkSectorWeight) * (BencharkSectorReturn - BencharkTotalReturn)

Is there a way of using BF to calculate sector allocation in the Exact Method? I was surprised the method used BHB for calculating sector allocation, since I've read many posts in this forum where it's told that it's not the correct way.

Regards,
Tomas

AA = (PortfolioSectorWeight - BenchmarkSectorWeight) * (BencharkSectorReturn - BencharkTotalReturn)

This formula makes complete sense.

Look at it this way, if you were trying to analyze whether overweighting or under-weighting a sector makes sense or not, just think... if the sector give higher return than overall benchmark then it makes sense to overweigh and vici-versa (remember benchmark total return is weighted average return of constituent sectors).

Plz feel free to ask for any clarifications.

Regards,
Sharad