Implement Alabama Unemployment Insurance (al_ui) (ref #8282) - #8283
Implement Alabama Unemployment Insurance (al_ui) (ref #8282)#8283daphnehanse11 wants to merge 13 commits into
Conversation
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #8283 +/- ##
=========================================
Coverage 100.00% 100.00%
=========================================
Files 3 12 +9
Lines 63 158 +95
Branches 3 0 -3
=========================================
+ Hits 63 158 +95
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
Program Review — PR #8283 (Implement Alabama Unemployment Insurance,
|
| Check | Result |
|---|---|
| Regulatory Accuracy | Correct — WBA/round-half-down/$275/3-prong eligibility/duration all match statute; no reinvented variables; documentation/pinpoint-citation items only |
| Reference Quality | 0 missing refs, format PASS; broken article-4 URLs + pinpoint subsections to fix; act-number to reconcile |
| Code Patterns | 0 critical / 3 should / 4 suggestion — zero hardcoded values, correct Person/YEAR entities, complete programs.yaml; stray lessons file + agency field |
| Test Coverage | Strong (round-half-down tie, $44.50, $275, 3 prongs, MBA both caps, partial boundaries, 11 integration); gap: middle duration brackets 16/17/18 |
| Source Audit | $275 / $44.50 / 1.5× / 2-quarter / 14-20-week scale / round-half-down all confirmed vs statute + BRR + USDOL |
| CI Status | Passing |
Review Severity: COMMENT
A correct, thorough, cleanly-parameterized new program with an unusually complete test suite and no blocking defects. Before marking ready: fix the broken reference URLs (item 1), drop the stray lessons/agent-lessons.md (item 2), align the programs.yaml agency field (item 3), add the middle-duration-bracket tests, and rebase off the 1689-commit lag.
Next Steps
To auto-apply the actionable items: /fix-pr 8283
Review generated with Claude Code via /review-program
|
Addressed the review items and rebased onto current main (branch is now main + 9 commits):
Alabama suite passes 391/391 locally (includes the 10 new bracket cases). |
Re-review — items addressed, and a correction to my own prior findingThanks for the fixes. First, an important retraction of the top item from my earlier review: ❌→✅ The Everything else I flagged is resolved:
Two minor residuals (non-blocking, optional):
Net: the implementation was already regulatorily correct (WBA/round-half-down/$275/3-prong eligibility/duration all verified), and this round removes the scratch file, aligns the agency field, and closes the duration-bracket test gap. LGTM once CI settles green (still running) and the branch is rebased off its main lag. Re-review via /review-program |
Adds the Alabama Unemployment Compensation program (Code of Alabama Title 25, Chapter 4) with monetary eligibility, weekly benefit amount, sliding-scale duration tied to state unemployment rate, dual-cap maximum benefit, partial-week earnings disregard, and the 1-week waiting period. All parameters effective 2020-01-01 per Act 2019-204. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- al_ui.py: fix partial-benefit fallback (was returning full WBA when
weekly_earnings >= WBA; should return 0 per "deemed employed" rule
in § 25-4-72 / USDOL Tbl 3-8). Now uses al_ui_partial_weekly_benefit
directly, which already handles this case correctly.
- Fix 5 BRR PDF page anchors (printed page vs file page; off by 3):
eligibility/{bpw_to_hqw_multiplier,quarters_with_wages}.yaml,
wba/{max,min_threshold}.yaml, waiting_weeks.yaml.
- Replace Ala. Admin. Code 480-4-3-.11 citation with statute § 25-4-73
for the 1/3 WBA partial disregard (480-4-3-.11 is procedural).
- Add al_ui entry to programs.yaml registry.
- Add test case for weekly_earnings >= WBA returning 0 (locks in the
al_ui.py fix).
- Fix two parameter description verbs ("requires" → "sets").
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Round 2 review identified that the "two quarters of your base period" sentence is on file page 7 of the BRR handbook (printed page 4), not page 8. Page 8 has the related "two highest base period quarters" phrasing, which supports the value but is less direct. Tightening the anchor for precision. Round 2 found 0 critical issues otherwise — all Round 1 fixes verified correct, no regressions. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Remove stray lessons/agent-lessons.md - Set programs.yaml agency to State per convention - Add just-below/just-at boundary tests for the 16/17/18-week duration brackets (Cases 5-14) - Document intentional modeling choices: partial weekly benefit applied to every payable week, and the statutory WBA/MBA formulas as authoritative over the BRR handbook's simplified gloss - Justia URLs and pinpoint cites verified correct as-is (Article 4; (b)(5)/(b)(2)) and left unchanged Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
|
@DTrim99 CI diagnosis for the one red check — this is runner memory topology, not a test failure. Evidence:
Interpretation: Proposed fix (happy to push either, but it's a CI-topology call so flagging first):
The code itself is unchanged since your approval. |
Program Review — PR #8283: Implement Alabama Unemployment Insurance (
|
| # | Document | URL | Status |
|---|---|---|---|
| 1 | AL DOL, UC Benefit Rights & Responsibilities Handbook (23 pp.) | https://labor.alabama.gov/docs/guides/uc_brr.pdf | Downloaded and audited |
| 2 | U.S. DOL, Comparison of State UI Laws 2023 — Monetary Entitlement (~90 pp.) | https://oui.doleta.gov/unemploy/pdf/uilawcompar/2023/monetary.pdf | Downloaded and audited |
Statute pages (Ala. Code §§ 25-4-72/-73/-74/-77 on law.justia.com, and Ala. Admin. Code 480-4-3) are bot-walled (HTTP 403 / JS wall) to automated fetch. Statute text was independently verified via the law.onecle.com and FindLaw mirrors of the same codification and Justia's indexed text; the cited Justia URLs resolve to the correct sections.
Branch Status
⚠ The PR branch is 9 commits ahead / 29 commits behind main. A rebase (or merge from main) is advised before merging. Branch staleness did not affect any finding below — all findings are in files new to this PR.
Critical (Must Fix)
C1. Duration bracket boundaries over-grant weeks — statute uses completed 0.5-point increments with the 20-week cap at ≥ 9.5% (VERIFIED)
parameters/gov/states/al/dol/unemployment_insurance/mba/duration_weeks.yaml encodes thresholds 0 → 14, 0.0651 → 15, 0.0701 → 16, …, 0.0901 → 20, i.e. any exceedance of a half-point grants the next week and the 20-week cap is reached at 9.01%.
Ala. Code § 25-4-74(a) (as amended by Act 2019-204) provides, verbatim:
"…the lesser of 14 times his or her weekly benefit amount, if the state's average unemployment rate is at or below 6.5 percent, with an additional weekly benefit amount added for each 0.5 percent increase in the state's average unemployment rate above 6.5 percent up to a maximum of 20 times his or her weekly benefit amount if the state's average unemployment rate equals or exceeds 9.5 percent…"
A week accrues per completed 0.5-point increment: weeks = 14 + floor((UR − 6.5) / 0.5), capped at 20 at ≥ 9.5%. The arithmetic check is decisive: 14 + 6 increments = 20, and six completed 0.5-point steps above 6.5% land exactly at 9.5% — the statute's cap trigger. Under the PR's brackets the cap is reached at 9.01%, leaving the statutory "equals or exceeds 9.5 percent" language with no possible role. The repo's brackets over-grant exactly one week everywhere strictly between half-points (6.6% → 15 vs statutory 14; 9.1% → 20 vs statutory 19); the two readings agree only at exact half-points.
Two internal inconsistencies confirm this:
- The YAML's own comment on the top bracket ("statute caps at 20 weeks for AAU >= 9.5%") contradicts its 0.0901 threshold.
- Every located secondary paraphrase (e.g. Alabama Daily News, Jan 7 2020) tracks the floor wording; advocacy and tracker sources (Alabama Arise 2021, CBPP, FGA) treat Alabama as a 14-week state throughout the pandemic, which is inconsistent with the PR's brackets for the Q3-2020 window.
Tests bake in the wrong reading: tests/.../al_ui_max_weeks.yaml Cases 2, 6, 8, 10, 12, 13 (6.51 → 15 … 9.01 → 20) must be rewritten to the floor schedule (e.g. 6.6 → 14, 7.0 → 15, 9.4 → 19, 9.5 → 20).
Blast-radius caveat: at recent Alabama unemployment levels the model output is 14 weeks under either reading for every modeled year (all January values 2020–2026 are far below 6.5%), so no current baseline al_ui output changes. This is a boundary-correctness fix, not a live output bug — but the encoded schedule is the load-bearing statutory rule and must be correct.
Fix: thresholds 0.07 / 0.075 / 0.08 / 0.085 / 0.09 / 0.095 (bracket lower-bound semantics already give ≥; no .xx01 fudge needed), corrected top-bracket comment, and rewritten tests. No change needed to the 14/20 endpoints or to al_ui_maximum_benefit_amount.py.
C2. state_unemployment_rate.yaml values match no published FRED/BLS vintage, and the 2025-06 – 2026-04 tail is extrapolated data presented as sourced (VERIFIED)
The file cites FRED ALUR (https://fred.stlouisfed.org/series/ALUR) and BLS LAUS, but independent re-verification against current FRED data and ALFRED archival vintages (first releases and post-benchmark vintages, 2020–2025) found:
- 46 of the 66 observable months (2020-01 … 2025-06) match no value in the PR-contemporaneous mid-2025 vintage (max deviation 0.7 pp, e.g. 2020-05: repo 11.1 vs 10.4; 2020-12: repo 3.8 vs 4.5 — no vintage ever published Dec 2020 below 3.9).
- 2020–2022 match essentially no published vintage at all (3 stray agreements in 36 months); 2023 – mid-2025 loosely tracks the mid-2025 vintage with ±0.1–0.2 pp transcription/smoothing errors.
- The 10–11 tail months (2025-06/07 through 2026-04, flat 3.4–3.5%) were unobservable when the PR was authored — they are forward-fill projections carried under FRED/BLS citations, and they overshoot the actual outturn (2.7–2.9%) by up to 0.8 pp.
This is a non-corroborating reference: the cited source cannot produce the encoded values, and unobservable future months are presented as sourced data with no projection note. (This finding subsumes the reference validator's independent flag of the same issue.)
Behavioral caveat: zero effect on current outputs — al_ui_max_weeks reads only the January value of each benefit year, and every January value on every vintage is far below 6.5%, so all modeled years give 14 weeks regardless.
Fix: re-pull the ALUR seasonally-adjusted series, state the retrieval/vintage date in the reference title (LAUS is re-benchmarked annually, so a bare series link is not reproducible), correct 2020–2025 to the pulled vintage, and either drop the post-observation months or explicitly label them as an assumption. Also note the series ends at 2026-04, after which the parameter silently carries the last value forward indefinitely.
C3. Incorrect statutory pin-cite: § 25-4-77(a)(4) should be § 25-4-77(a)(6)
eligibility/bpw_to_hqw_multiplier.yaml and eligibility/quarters_with_wages.yaml (and test-file comments) cite the qualifying-wage rule as "Code of Alabama § 25-4-77(a)(4)" / "(a)(4)(a)". In the verified codification (stable since at least 2006; confirmed via onecle/Justia), the "equal to or exceeding one and one-half times…" qualifying-wage requirement is subdivision (a)(6), and no "(a)(4)(a)" sub-paragraph containing the 1.5× rule exists. All encoded values (1.5 multiplier, 2 quarters, inclusive ≥) are correct — only the pin-cite is wrong. Trivial fix: correct the subsection references.
C4. NELP citation is the wrong vintage and contradicts an encoded value
partial/disregard_rate.yaml and al_ui_partial_weekly_benefit.py cite NELP, "New Alabama Unemployment Insurance Law Makes Work Pay" (https://www.nelp.org/new-alabama-unemployment-insurance-law-makes-work-pay/) — a May 2015 article about the 2015 law. It does corroborate the one-third disregard, but the same article states the maximum weekly benefit is "$265," the pre-Act-2019-204 figure that contradicts the PR's (correct) $275 encoding. A stale secondary advocacy source is standing in near the primary statute (§ 25-4-73). Fix: keep § 25-4-73 as the corroborating citation of record for the 1/3 disregard; if NELP is retained, annotate it ("2015 article — describes the one-third disregard; its $265 max is pre-2019 and superseded").
Should Address
S1. Unemployment-rate measure: January single-month value vs the statutory prior-Q3 three-month average — undocumented approximation
Ala. Code § 25-4-74(d) defines the governing rate as "the average of the three months for the most recent third calendar quarter of the seasonably adjusted statewide unemployment rate as published by the Alabama Department of Labor." The model instead keys duration off a monthly parameter that, at YEAR definition period, resolves to the January value of the benefit year. The in-code comment in al_ui_max_weeks.py documents the January-value framework behavior but nowhere states that the statute prescribes a prior-Q3 average — the approximation is undocumented as an approximation. Materiality: for 2022+ both measures sit far below 6.5% (14 weeks either way). It matters in the COVID window — the Q3-2020 SA average (~6.6%, current vintage 6.57%) would govern late-2020/2021 claims, and this approximation compounds with C1 exactly there (repo brackets → 15 weeks; statutory floor reading → 14). At minimum, document the simplification in the variable docstring and the parameter file; ideally move toward the statutory Q3-average measure.
S2. al_ui is not connected to household net income (orphaned benefit)
Nothing adds al_ui to unemployment_compensation, spm_unit_benefits, or any net-income aggregate — the same is true of the merged pa_uc, so this is a pattern-level limitation, not a regression. With economy: false / household: true it is a household-calculator output only. Confirm this is intentional and state it in the PR description so it is not mistaken for a wired-in benefit; consider a follow-up issue for wiring state UI into net income.
S3. programs.yaml verified_years: "2020-2025" overstates test evidence
All 78 test cases run at period: 2024. Parameters carry 2020-01-01 values and monthly ALUR through 2026-04, but 2020–2023 and 2025 have no test coverage. Either add at least one integration test at an early year and one at 2025, or narrow verified_years to what is actually tested.
S4. Input variables default to zero — the program is inert in microsimulation
The five inputs (al_ui_high_quarter_wages, al_ui_second_high_quarter_wages, al_ui_base_period_wages, al_ui_quarters_with_wages, al_ui_weekly_earnings) have default_value = 0 and no imputation, so in any dataset run the whole AL population is monetarily ineligible and al_ui contributes $0 (same as the PA UC inputs). Expected for a first-cut encoding; flag it so reviewers know the program only activates in household calculations with explicit inputs.
S5. Not-modeled provisions are undocumented — notably the +5 training weeks (§ 25-4-74(f))
§ 25-4-74(f) and the BRR Handbook ("Additional Training Benefits": claims on/after Jan 1 2020 get five additional weeks, 5 × WBA added to the MBA) are unmodeled and unmentioned anywhere in the diff. Omission is defensible (requires a training-enrollment input the model lacks), but it should be documented. Unlike pa_uc, al_ui.py carries no "Not modeled:" docstring — add one listing the exclusions (non-monetary eligibility, work search, training extension, experience rating, CWC).
S6. MBA rounding: np.round (banker's rounding) vs the claimed "nearest dollar," with no .50 tie test
al_ui_maximum_benefit_amount.py uses plain np.round, which rounds half-to-even, while its comment claims statutory nearest-dollar rounding; no statutory rounding provision for the MBA itself was located (the nearest-dollar clauses in §§ 25-4-72/-73 govern the WBA and partial benefit). The MBA tests pin 1,250.75 → 1,251 and 1,250.25 → 1,250 — neither is a tie. Add a case landing exactly on X.50 (e.g. BPW = 5,002 → 0.25 × BPW = 1,250.50) to pin the intended direction, and reconcile the comment with the implementation.
S7. Test boundary gaps at the clamps
(a) No case at weeks_unemployed = 15 (the first week where the max_weeks cap binds after the waiting-week subtraction) — the upper clamp of clip() is only exercised at 20. (b) No annual al_ui case with MBA one dollar above the annual payout (the just-below crossover direction of the min_()). (c) No explicit sub-6.5% override (e.g. state_unemployment_rate: 0.0) pinning the floor bracket — note these bracket tests must be rewritten anyway under C1.
S8. Reference hygiene
bpw_to_hqw_multiplier.yamlcitesuc_brr.pdf#page=8, but the BRR Handbook contains no 1.5× multiplier, no WBA formula, and no rounding rule — that anchor does not corroborate the value (the statute and USDOL Table 3-2 citations on the same parameter do the real work). Remove or retarget it.- Demote secondary sources: alabamaretail.org (Act 2019-204 summary) confirms the 14–20 float and $265 → $275 but contains none of the bracket thresholds; the statute should be the source of record on
duration_weeks.yamlandwba/max.yaml. - Confirm the remaining
#page=anchors land on the cited tables (the USDOL PDF's internal numbering is 3-x; anchors must be file pages). waiting_weeks.yamllinks the Ala. Admin. Code 480-4-3 chapter index; deep-link the specific rule (the partial-benefit file already deep-links 480-4-3-.11).
Suggestions
- Per-week partial-earnings assumption —
al_ui.pyapplies oneal_ui_partial_weekly_benefituniformly across all payable weeks (reduces to full WBA at zero earnings). Reasonable and inline-commented; confirm intended semantics for the annual total. al_ui_weekly_earningsis a YEAR-period variable holding a weekly figure (mirrors PA'spa_uc_gross_weekly_earnings). Consider a clarifying note; low priority.- Move defensive regulatory rationale to docstrings — the multi-line "authoritative over the BRR handbook… should not be corrected" comments in
al_ui_unrounded_wba.py/al_ui_maximum_benefit_amount.pywould sit better in the class docstring, as PA UC does. - Description verb —
state_unemployment_rate.yaml"Alabama uses this…" is on the borderline of the allowed-verb list; minor style nit. single_amountbracket →intcast —al_ui_max_weeksisvalue_type = intover a bracket returning float amounts (14–20); confirm the cast is clean. Low risk.- Test polish — pin a
.50tie for the partial benefit; assert an intermediate (e.g.al_ui_weekly_benefit_amount: 0) in the non-AL case sodefined_forzeroing is distinguishable from zeroed inputs; add a negative wage-input case; assertal_ui_unrounded_wbain the WBA boundary cases. - Future backdating notes — max WBA was $265 before 2020-01-01 and the waiting week dates to claims on/after 2012-08-01; the 2020-01-01 parameter starts are correct for the declared window, but a backdating pass must add the earlier tiers.
- Citation-year maintenance — the USDOL Comparison is the 2023 edition; fine for the constants (all unchanged since Act 2019-204), but if Alabama amends any amount the citation year must advance.
Verified-correct notes for future maintainers (do not "fix" these): the WBA half-down rounding np.ceil(x − 0.5) exactly implements § 25-4-72(b)'s ">$.50 up, ≤$.50 down" rule — USDOL Table 3-5's "Higher $" label is imprecise; and Table 3-11's "Lesser of 14 x WBA" is point-in-time shorthand — the statute scales the WBA-multiple prong 14→20 with the unemployment rate, exactly as al_ui_maximum_benefit_amount.py encodes.
Source Audit Summary
Both cited PDFs were downloaded and audited page-by-page; statute text was verified via codification mirrors (Justia pages bot-walled).
| Audit area | Values confirmed | Confirmed mismatches | Notes |
|---|---|---|---|
| WBA & monetary eligibility (formula, rounding, $44.50 strict floor, $275 max, 1.5× HQW, 2 quarters, cap-then-gate) | 9 | 1 | Mismatch is citation-only: § 25-4-77(a)(4) → (a)(6) (C3); all values and rounding verified correct |
| Duration / MBA / partial / waiting week / UR series | 10 | 2 | Duration bracket boundaries (C1); UR series vintage/extrapolation (C2). Q3-average measure logged as an undocumented approximation (S1), not a value mismatch |
| Total | 19 | 3 | Both substantive mismatches independently re-verified by dedicated verifier passes (verdict: CONFIRMED) |
Validation Summary
| Validator | Result |
|---|---|
| Regulatory review (statute + USDOL cross-check) | 0 critical, 3 should-address, 3 suggestions |
| Reference validation | 2 critical (1 merged into C2; 1 kept as C4), 5 should-address, 4 suggestions |
| Code pattern audit (vs merged PA UC pattern) | 0 critical, 3 should-address, 5 suggestions |
| Test coverage review (~78 cases, 8 files) | 0 critical, 4 should-address, 6 suggestions |
| PDF audit — WBA/eligibility | 9 matches, 1 citation-only mismatch |
| PDF audit — duration/MBA/partial/UR | 10 matches, 3 mismatches |
| Verifier — duration brackets (codepath 1) | CONFIRMED (statutory floor reading; repo over-grants) |
| Verifier — UR series vintage (codepath 2) | CONFIRMED (no vintage match; tail extrapolated) |
| CI Status | ⚠ Failing: states-shard-1 — diagnosed on the PR thread as a runner OOM infrastructure issue, with a proposed workflow fix (test batches 16 → 20). All other checks green. Not a model-correctness finding, but merge requires resolving it (apply the workflow batching bump or admin-merge). |
Review Severity
REQUEST_CHANGES — 4 critical, 8 should-address, 8 suggestions. The criticals are: the statutory duration-bracket schedule (C1, with tests to rewrite), the unemployment-rate data series and its citations (C2), and two citation defects (C3 trivial pin-cite; C4 stale contradictory secondary source). None of C1/C2 changes any current modeled output (all recent Alabama rates yield 14 weeks either way), so the fixes are contained and low-risk.
Next Steps
Run /fix-pr 8283 to apply the findings above. Suggested order: C3 (one-line cite fix) → C1 (brackets + comment + test rewrite) → C2 (re-pull ALUR with vintage note) → C4 (annotate/replace NELP) → S1–S8 documentation and test additions. Also rebase onto main (29 commits behind) and resolve the states-shard-1 CI infra failure via the proposed workflow batching change.
This review was generated by the /review-program multi-agent pipeline (regulatory, reference, code-pattern, and test validators; page-level PDF audits; independent verifier passes on disputed findings).
- Correct duration_weeks brackets to the completed-increment reading of § 25-4-74(a) (14 + 1 week per completed 0.5pp above 6.5%, cap 20 at >= 9.5%); rewrite interior bracket tests to match - Replace state unemployment rate series with the FRED ALUR (seasonally adjusted) vintage, truncated at last published month; document the January-value approximation of the statutory prior-Q3 average - Fix § 25-4-77 pin-cites to (a)(6); drop stale NELP citation and non-corroborating anchors; add backdating breadcrumbs - Narrow programs.yaml verified_years to 2024; document unmodeled provisions and microsim-inert defaults in al_ui.py - Add edge tests: UR floor, week clamp, MBA min crossover, half-to-even ties, negative wages, non-AL zero WBA, unrounded WBA assertions - Bump states-shard test batches 16 -> 20 to fix runner OOM in CI Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Fixes Applied (from Program Review)Commit 9139969 addresses all actionable findings from the review above. Critical
Should Address
CI
Skipped (deliberate, no code change)
Verification
🤖 Applied via |
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Each distinct dotted-path parameter override deep-copies and caches a full tax-benefit system for the whole test subprocess, and states batching runs all AL files in one subprocess at every batch count. The 14-value state_unemployment_rate sweep in al_ui_max_weeks.yaml (~6.5+ GB resident) deterministically killed the runner at 16, 20, 24, and 32 batches. Keep four probes (floor threshold, completed-increment interior, exact step, cap) plus the no-override baseline, and restore the workflow's original 16-batch layout since the batching bumps were addressing a symptom. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
CI green — root cause of the states-shard-1 failures found and fixedAll 33 checks now pass at d1d446b. The recurring Commit d1d446b trims the sweep to four probe values (floor threshold, completed-increment interior, exact step, cap) plus the natural baseline, adds a memory-note comment to keep sweeps from returning, and restores 🤖 Diagnosed and applied via Claude Code |
2 participants
Summary
Implements the Alabama Unemployment Insurance program (Code of Alabama Title 25, Chapter 4).
Closes #8282.
Regulatory Authority
Monetary Eligibility (§ 25-4-77(a)(4))
A claimant is monetarily eligible if ALL of the following are true:
Weekly Benefit Amount (§ 25-4-72(b))
np.ceil(x - 0.5))Maximum Benefit Amount and Duration (§ 25-4-74(a))
Partial-Benefit Provisions (§ 25-4-72; NELP 2015 reform)
Waiting Week
Requirements Coverage
eligibility/quarters_with_wages.yamlal_ui_monetarily_eligible.pyal_ui_monetarily_eligible.yamleligibility/bpw_to_hqw_multiplier.yamlal_ui_monetarily_eligible.pyal_ui_monetarily_eligible.yamlwba/min_threshold.yamlal_ui_monetarily_eligible.py,al_ui_weekly_benefit_amount.pyal_ui_weekly_benefit_amount.yamlal_ui_unrounded_wba.pyal_ui_unrounded_wba.yamlal_ui_weekly_benefit_amount.pyal_ui_weekly_benefit_amount.yamlwba/min_threshold.yamlal_ui_weekly_benefit_amount.pyal_ui_weekly_benefit_amount.yamlwba/max.yamlal_ui_weekly_benefit_amount.pyal_ui_weekly_benefit_amount.yamlmba/duration_weeks.yaml,state_unemployment_rate.yamlal_ui_max_weeks.pyal_ui_max_weeks.yamlmba/bpw_fraction.yamlal_ui_maximum_benefit_amount.pyal_ui_maximum_benefit_amount.yamlal_ui_maximum_benefit_amount.pyal_ui_maximum_benefit_amount.yamlal_ui_partial_weekly_benefit.pyal_ui_partial_weekly_benefit.yamlpartial/disregard_rate.yamlal_ui_partial_weekly_benefit.pyal_ui_partial_weekly_benefit.yamlwaiting_weeks.yamlal_ui.pyal_ui.yaml,integration.yamlNot Modeled
Historical Notes
All AL UI parameters in this PR start at 2020-01-01, matching the effective date of Act 2019-204 which introduced:
The 1/3 partial-earnings disregard has been in effect since 2015-06-13 (prior was a flat $15) — using the 2020-01-01 effective date in the parameter is conservative (i.e., the rule was already in effect by then).
State unemployment rate values in
state_unemployment_rate.yamlare approximate decimal-form FRED ALUR series values for 2020-01 through 2026-04. Reviewers should verify these against the authoritative FRED series before merge.Test Coverage
Files Added
Modified:
policyengine_us/variables/gov/states/unemployment_compensation.py— addedal_uito theaddslist so it flows into the federalunemployment_compensationaggregator.