The problem is that the concept isn't precise enough. You'll find striking differences within a given "racial group" once you break it down into more useful descriptors. See, for example:
https://en.wikipedia.org/wiki/List_of_ethnic_groups_in_the_U...Look at Native Americans, where the median household income is almost double for some subgroups compared to others. For Asians, the income difference is more than double between Indians and Bangladeshis. American Turks seem to do extremely well, while I expect the numbers for German Turks would be a fair bit lower.
So if you're trying to analyze things, whether for medicine (blood pressure), or sociology (impacts of racism, benefits of economic intervention), racial groups are a pretty poor categorization, vulnerable to Simpson's paradox.
One classic conservative argument against reparations is that it lumps in rich Nigerian immigrants with poor African-American descendants of slavery. Seems like their logic could actually be used to support this change.