We’re consulting with a non-software vendor client who is enhancing their performance measurement system, which includes transaction-based attribution. During a recent discussion they mentioned that they calculate attribution daily and then link these results. Seems like a lot of work to me. Plus, a great deal of processing time and cost associated with it.
Intuitively one would expect that if you did daily transaction-based attribution your results would be better than if you did monthly: you’d be wrong. Transaction-based attribution can be done quite effectively monthly: you should get essentially the same results, save for perhaps some rounding issues.
The story is different with holdings-based attribution: if you want to increase its accuracy, daily is ideal.
By using modified Dietz to capture the returns and weighting the cash flows (buys and sells) the same way for securities/sectors as you would at the portfolio level, you’ve got all you need to ensure accuracy.
To add credibility to my claim I contacted my friend Carl Bacon, chairman of Statpro, who concurred (always nice when we agree!). And so, don’t make this stuff harder than it needs to be!