Why use various proxies at all? Most of the group differences seem to largely share poverty as a primary root factor, and this impacts plenty of sub-groups within the supposedly advantaged groups as well. So why are we not simply addressing that directly?

A refrain I've been seeing a lot lately is that the outcome of a system is its purpose. Perhaps that reasoning should be equally applied to an analysis of the implementation choices made by various social programs and the solutions espoused by various political movements.

the fun part here is seeing interesting tradeoffs like a wealthy person of <Race-preferred> chosen over a poor person of <Race-tarnished> for some kind of wealth producing choice. Making a wealthy member of that group wealthier will pull up the average, but IMO the just choice is to help the lagging human regardless of their race.

It leads to weird outcomes too like the deaths of despair in Appalachia being "non news" because it largely covers a group we deem as "privileged".

A big part of the issue is reductionistic / simplistic thinking that wants to reduce the cognitive load of decision making by setting up only a few binary gates instead of having to do complex analysis on a case by case basis. This is normal for human brains to try and do.

Why? I presume that people don’t look for second and third order effects. They see disproportionate numbers of people of color struggling, they blame the first order grouping as the cause.

Statistics is complicated. People reject even straightforward theories with obvious scientific bases. It makes sense.