I am tracking the risk of portfolios using a standard 2-asset benchmark (S&P500 / Agg bond) and I want to throw a flag if the risk of a portfolio goes outside of a certain range relative to that benchmark. I am looking for resources that address whether or not to allow the weights to the risk benchmark to drift over time.
As a sample case, the benchmark is 70% S&P500 / 30% Agg bond
The simplest thing, of course, is to create a series of returns such that the benchmark return in every month = 0.7 * return of S&P500 + 0.3*return of Agg Bond. This amounts to assuming a monthly rebalance of the benchmark.
Alternatively, we can form the 70/30 benchmark and then allow the weights to drift over some extended period, calculate monthly returns of the drifting benchmark, and calculate volatility from that. The period over which you allow the portfolio to drift would most naturally be equal to the length of historical period over which you calculate volatility (but these could be different).
In my experience, people follow the simple solution but I can see some merits to the second. Any suggestions?