DE jnt 2025 85Kintrec -2848Kstcg 98Kltcg 10Kswages

[feenberg]

See http://taxsim.nber.org/out2psl/916

TA chooses married filing separate combined return. Both possibilities are attached and separate combined looks better. PE result is close to the Joint filing tax. Maybe PE chose wrong? Maybe property income wasn't split in half?

[PavelMakarchuk]

Tracked upstream: PolicyEngine/policyengine-us#8501. PE actually picks combined filing (status 4) here — same as TaxAct — but de_pension_exclusion_income sums net capital_gains (including stcg losses), which makes the eligible income strongly negative. de_pension_exclusion.py:44 then returns min($12,500, negative) = negative, and de_subtractions = max(0, …) = $0 — so the 75-year-old loses the $12,500 exclusion they should get. Removing the cap losses from the same case reproduces $12,500 ✓. Once #8501 lands, the DE tax should come within ~$10 of TaxAct's $2,851.

[PavelMakarchuk]

Correcting my earlier framing — the upstream fix (PolicyEngine/policyengine-us#8501, merged) addresses a different mismatch than I originally described.

What the fix does: floors capital_gains at zero inside the DE pension-exclusion basket so a large net capital loss can't drive de_pension_exclusion strongly negative — which previously zeroed out every DE subtraction on the return (US-government interest, IRA adjustments, etc.) via de_subtractions = max(0, sum). For this issue's repro, there are no other DE subtractions, so the structural fix produces $0 visible tax change — PE still returns ~$3,547.

What the fix does not do: close the PE-vs-TaxAct $688 gap. That delta reflects TaxAct applying the full $12,500 pension exclusion despite the net capital loss. DE PIT-RES Line 6 instructions describe the exclusion-eligible basket as "capital gains net of capital losses."