Sometimes because if you add 'race' to the data then clusters appear that align with that extra data. Of course it doesn't mean that 'race' is a causal factor, it can be that it correlates with some other data that is not included such as poverty.
For many outcomes (including medical outcomes) the differences do not go away when conditioning on wealth/income, so if there's a separate "root" causal factor, we have not yet found it.
It's too bad we don't in the US break down Black population statistics into those who are descendants of slaves vs. more recent immigrants - the differences in life outcomes would probably be significant, even at the same income level.
