Your analysis appears to use the price-only version of the CAPE ratio, and you don't discuss a total-return CAPE as advocated by Jeremy Siegel in the Financial Analysts Journal and also by Shiller ("The Many Colors of CAPE"). The use of a total return CAPE is motivated explicitly by changes in accounting rules since 1990, which seems to fit your break-date analysis. Are your empirical results affected if you use a TR CAPE instead of a price-only CAPE?