One thing I will discuss at this year's Spaulding Group PMAR (Performance Measurement, Attribution & Risk) conferences is the impact of benchmark changes to the attribution residual. While we recognize that the holdings-based model suffers from residuals when trades occur, one can easily overlook the contribution that can arise from benchmark turnover.
While doing my research, I initially set out to avoid any months that had any turnover in the index (I'm using, yes you guessed it!, the S&P 500). But finding months like this can be a challenge, and so I decided to ignore this rule. I had temporarily forgotten the basis for my initial plan (to not have to worry about the turnover), but was quickly reminded when I saw a residual with both the holdings and transaction-based approaches, when there was no activity in the period. How could this occur? Only one answer: turnover in the benchmark.
This is an important topic to address with vendors, if you're engaged in an attribution software search, and are looking for a transaction-based model: make sure the system is sensitive to benchmark turnover! Otherwise, you'll occasionally see a residual appear.