Do this for two sample periods: 7/1926-12/1963, and 1/1964- 08/2017 (in each case the beta should be computed using the whole subsample, although in practice we use shorter time periods e. g. 5 years). How well does the CAPM work in the first sample? What happens in the second sample?