Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (752fae4) to head (c2c5457).
⚠️ Report is 40 commits behind head on main.

@@            Coverage Diff            @@
##              main     #7753   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            2         1    -1     
  Lines           13        18    +5     
=========================================
+ Hits            13        18    +5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Oh I'm so stuck here. Concerned about this:

  This is a circularity bug: is_filer depends on aca_ptc (via eligible_for_refundable_credits), and aca_ptc depends on county. The constraint evaluates is_filer at state level where aca_ptc = 0, so is_filer = False. But the actual county-level aca_ptc > 0 would make is_filer = True.

Hi @juaristi22 — we've been working through a related issue in policyengine-us-data's calibration pipeline and concluded that the causal direction here should be reversed.

In reality, ACA marketplace enrollment (receiving APTC) creates a legal obligation to file (IRC § 36B reconciliation). So rather than gating aca_ptc on is_filer, we should make ACA enrollment imply filing.

PR #7782 adds takes_up_aca_if_eligible as an OR condition in tax_unit_is_filer, so ACA enrollees are always filers. This makes the * is_filer gate proposed here unnecessary — enrollees will always pass it.

See #7781 for the full discussion. Would love your thoughts on this approach.

After further investigation, we believe the current behavior on main is correct — aca_ptc should NOT be gated on is_filer.

The ACA Premium Tax Credit is received as APTC (monthly premium reductions) at enrollment time, before any tax filing happens. Filing is required to reconcile, but the benefit flows regardless. Gating on is_filer would incorrectly zero out PTC for enrollees who received the benefit but haven't filed.

The takes_up_aca_if_eligible draw (67.2% rate from KFF data) already handles the participation question. Adding * is_filer would create a double filter with the wrong causal direction — filing doesn't gate the benefit, enrollment does.

Apologies for the earlier comment suggesting our PR #7782 as an alternative — we got confused ourselves and are closing that one too. See below.