This is many things:

What you must remember though, is that the kind of microdata to required to do this is either non-existent or paywalled. In many cases, there are policy hurdles to getting the data. The statistical realities detailed within are generally true even if not numerically accurate.

Given that, you should and can use numbers within DiD as degrees of belief rather than an actual proportion. Generally speaking, any Bayesian statements you make based on the data on this website will likely be true.