Does `taxsimtest` apply the Oregon kicker only for 2021 (and not 2023's 44.28% or 2025's 9.86%)?

[PavelMakarchuk]

Diagnosing Oregon in the eCPS comparison (#1059): OR is the last large unexplained 2021 state (24% match at every tolerance). The mechanism is the Oregon kicker (Or. Const. Art. IX §14; OR-40 refundable credit = kicker% × prior-year tax before credits):

• TY2021: taxsimtest applies the 17.341% kicker, proxying the prior-year (2020) tax with the current-year staxbc — verified exactly: siitax = staxbc × (1 − 0.17341) on clean records (e.g. staxbc 375,869.50 → siitax 310,690). PE computes a $0 kicker (its or_tax_before_credits_in_prior_year input has no TAXSIM column to feed it), so PE runs ~21% high on essentially every 2021 OR record.
• TY2023: Oregon's certified kicker was 44.28% — but the binary's implied credit rate is only ~5% (ordinary credits): no kicker applied.
• TY2025: certified 9.863% — also not applied.

Year	Kicker (certified)	binary applies?
2021	17.341%	yes (current-year proxy)
2023	44.28%	no
2025	9.863%	no

Two questions:

1. Is the 2021-only application intentional, or should the 2023/2025 kickers (44.28% is a very large credit — every 2023 OR-40 has it) be applied as well?
2. Is the current-year-staxbc proxy for the prior-year base the intended convention? (PE would mirror the same proxy for parity once the year coverage is settled — we've held off implementing so the two engines don't leapfrog each other.)

Happy to send txpydata records. PE currently applies no kicker in any year (prior-year input unavailable from TAXSIM input), so the engines coincidentally agree in 2023/2025 — both omitting a certified credit — and diverge ~21% of tax on every 2021 OR record.

[feenberg]

Please always send txpydata.csv record.

[PavelMakarchuk]

txpydata record (a clean OR isolation), with your option columns:

taxsimid,year,state,mstat,page,sage,depx,pwages,intrec,idtl,opt1,opt1v
900010,2021,38,1,40,0,0,100000,0,2,30,1

And an update that mostly answers my own question: re-testing with opt(30)=1 (your PSL-conformance mode), the binary applies no kicker — siitax = staxbc = 7,670.92 — which matches PolicyEngine exactly (PE's or_tax_before_credits_in_prior_year input has no TAXSIM column, so its kicker is $0). Default mode applies the 17.341% kicker in 2021 (7,670.92 → 6,340.71) but not 2023's 44.28% or 2025's 9.86%.

So under your testing protocol there's no OR discrepancy; the residual question is only about default mode: is 2021-only kicker coverage intentional (vs 2023/2025), and is the current-year-staxbc proxy for the prior-year base the intended convention there? Low priority — our comparisons will switch to opt(30)=1 to measure the mode you actually test.