Dynamic Beta

The phrase “Dynamic Beta” is broad and this paper describes statistical procedure for estimating regression coefficients in a way that allows for variation across relevant subsets of the data. For example, the time axis. I describe an algorithm to structure the search for variation in sets of coefficient estimates and discuss the example of a single stock versus a stock index. In the end, I suggest that a human analyst has an important role for someone who has relevant skill in pattern recognition and subject area expertise