This is a nice instructive example in causal inference with observational data. The authors aimed to find evidence for class-based disparities in academic career success between people who are first generation college graduates who got PhDs and those who have parents with graduate degrees. They did find differences in progression (e.g. getting tenure), but then also found differences in performance. From a footnote:
>Indeed, conditional on PhD institution and field, we do find that tenured academics who are first-gen college grads have fewer publications, fewer citations, and lower average journal impact factors than those with a parent with a non-PhD graduate degree (Appendix Table A6). But as discussed in the text, this research gap could be an outcome, not a cause, of the class gap in tenure institution type. This is why we do not run our baseline ``tenure anywhere'' or ``tenure at R1'' regressions controlling for research output: because those who are either non-tenured or tenured at non-R1s will have substantially less time, resources, and incentive to produce research
Naturally, both differences in progression and performance can be explained by having parents with more education (and implicitly, more wealth), but the issue of possible class bias (e.g. someone getting tenure because the committee is somehow influenced by the candidate having a mom who is also a professor) only makes sense if performance wasn’t mainly/largely driving progression. This is a complex element of the data generating process that cannot be distinguished using statistics on observational data without some very strong assumptions. The authors at least implicitly make these assumptions, which is typical to do in observational causal inference, but readers should also use their own reasoning and experience to decide for themselves.