One is the St Louis Fed study, which I found you and I have discussed briefly in the past. ;)
https://www.stlouisfed.org/~/media/files/pdfs/hfs/is-college...I’m only skimming quickly, this doesn’t fully back me up, but section III talks about some causal analysis, for example: “mediation tests causal relationships between variables. While the SCF is not longitudinal and thus a pure causal effect cannot be tested, we can be confident that reverse causality in the ordering of these variables is not possible.”
I have to be fair and note that there are also several comments about what is not causal as well as data that implies more causation than what their analysis reveals.
The references in this paper point to a couple of other papers that have “causal” in the title. Unfortunately I don’t have the time right now to read these again, but I recall seeing causal analysis during that previous discussion about the St. Louis Fed study, research that studied education outcomes and controlled for people with similar SES backgrounds, similar family history of education, etc., and found the apparent causal contributions of education.
BTW, there is absolutely a fuzzy gray line here. What I’m calling causation (social bias) may be what you’re calling correlation (education alone is not the entire causal reason for higher incomes - to your point that one cause is the people). Lest we get too stuck on terminology, I’m perfectly happy pointing out that the correlation is extremely high, and the statistical difference between with degree and without degree is surprisingly large.