> which is exactly the kind of meaningless result you'd expect from doing ~20 statistical tests over random numbers. (The paper lists hundreds of p-values.)

On page 20 they describe using the Westfall and Young method to adjust for multiple hypothesis testing and report adjusted and unadjusted p-values.

They also pre registered their analysis with depression as a primary outcome and breaking down by medical debt quartile as pre registered subgroup analysis.

In the highest quartile of medical debt group, 45.9% percent in the no debt forgiveness group had at least moderate depression vs 58.3% in the debt forgiveness group, p=0.002

Dismissing this as a meaningless result seems unfair given that they preregistered their analysis and adjusted for multiple comparisons.

I’m not a stats expert and haven’t spent much time with the paper, but I’d suggest checking the methods section before making claims.