Characteristic-varying take-up priors: EITC by qualifying-child count, SSI by age band

Max, 7/21, confirming the #469 direction: take-up is seeded Bernoulli from documented priors, and priors should vary by characteristics where the evidence supports it — the canonical case being EITC take-up by qualifying-child count (childless take-up is far below with-children take-up in the admin literature), and SSI by age band (the #470-class SSA age tables give the denominators' anchors). Pattern is program-generic: a prior table keyed by observable characteristics, documented with sources in the stage spec, seeded draws against it, and the corresponding registry count targets (EITC child-count decomposition targets already exist; SSI age-band targets arrive with the SSA age facts issue) reporting the post-calibration miss. No count-matching, no per-program assignment machinery — priors and targets only.

Read the #469 direction and the #470 SSA age tables first, then locate the stage spec, registry count targets, and existing EITC child-count decomposition targets. Done means characteristic-keyed priors are documented and seeded draws report post-calibration misses for EITC and SSI without count-matching or per-program assignment machinery.