Skip to content

Activate uprating for filing-status parameter children - #10825

Merged
MaxGhenis merged 10 commits into
mainfrom
fix-breakdown-parameter-uprating
Nov 16, 2026
Merged

MaxGhenis merged 10 commits into
mainfrom
fix-breakdown-parameter-uprating

Conversation

@MaxGhenis

@MaxGhenis MaxGhenis commented Nov 12, 2026 •

Copy link
Copy Markdown
Contributor

Activate uprating for the filing-status children of Colorado's CollegeInvest subtraction, the legacy personal exemption phaseout and the Biden FY2025 capital gains proposal.

Colorado's Department of Revenue (DOR) publishes the 2026 limits in its November 2026 bulletin, page 2:

Tax Year | Single Filers | Joint Filers
2026 | $26,200 | $39,200

Add those published limits and project subsequent years. Correct the Colorado statute reference from section 39-22-104(4)(j) to (4)(i), and replace the remaining Justia citations for CollegeInvest and the income tax rate with official legislature copies.

Correct the 2032 head-of-household exemption phaseout threshold from $421,050 to $421,200, following CBO's January 2025 forecast, CSV row CY2032,tp_pep_threshold_hoh,421200.0.

From 2026, Colorado filers who contribute more than $25,400 ($38,100 on a joint return) can subtract more and owe less state income tax. Congress keeps personal exemptions suspended, and the Biden proposal applies only when a user turns it on, so the other two repairs change only counterfactual reforms.

Sources: Colorado statute, section 39-22-104(4)(i)(II)(B)-(C), PDF pages 392–393, historical IRC 68(b), IRC 151(d)(3)(B), IRC 151(d)(5)(A), and Treasury's FY2025 proposal, printed page 80.

Invariants

  • Preserve authored values during uprating, including the corrected CBO threshold and the published Colorado 2026 limits.
  • Project Colorado caps to the nearest $100 and keep all nonjoint caps equal.
  • Round legacy exemption thresholds down to $50, or $25 for separate filers, preserving half the joint threshold after 2035.
  • Preserve the proposal's half-threshold for separate filers and equal thresholds for other statuses.
  • Keep exemptions suspended and the proposal inactive under current law.

The Python tests read built parameter values and metadata that household YAML tests cannot inspect. They reuse the existing system, check all 115 authored entries, and sample projected years through 2100 for rounding, growth and filing-status relationships.

Household YAML cases check Colorado caps, restored exemptions and proposal thresholds. After the merge with main, exemptions.yaml holds 26 cases: the 21 that #10854 added for 2017 through 2026 and this PR's five for 2032 through 2038.

Since #10854, the exemptions formula counts a fraction of a phaseout step as a full step, following IRC 151(d)(3)(B). Each restored-exemption case checks adjusted gross income $1 below its threshold, at it, $1 above it and one full step above it. The joint-filer case distinguishes rounding down from rounding to the nearest $50.

Validation

Pull request run 43944374859 at ec5403f3b4 passed all 37 jobs on GitHub's merge with main 29f81ccc36.

Under policyengine-core 3.36.1, the five changed test files run 180 cases, and all pass: 133 parameter, seven placement, 26 exemption, ten Colorado and four proposal cases. CI runs exemptions.yaml in the irs baseline shard, the Colorado file in a states shard, the proposal file in a contrib shard and the two Python files in the Rest Python shards. Ruff format and Ruff lint pass on the changed Python files, and git diff --check against main prints nothing.

Main changed none of the three parameter files and none of the chained CPI-U values behind the IRS index, so the merged tree projects the same 15 filing-status children to the values the tests expect.

Methodology choices

  • Under C.R.S. 39-22-104(4)(i)(II)(C), the Department of Higher Education calculates each year's change in tuition, room and board costs at state colleges, and the Department of Revenue may round the adjusted limits to the nearest $100. Anchor projections to DOR's published 2026 limits, and project later years with the existing IRS inflation index in place of that cost change, rounded to the nearest $100.
  • Preserve CBO's legacy forecasts through 2035, with the corrected 2032 head-of-household value. Project from the final forecast and apply historical IRC 68(b) rounding for counterfactual reforms.
  • Treasury specifies inflation adjustment after 2024 without choosing an index or rounding interval in the capital-gains subsection. Retain the existing IRS index and apply no rounding.
  • Retain the shared IRS index and its measurement window. Issue #9105 and PR #9621 track that window. PR #10783 added base-year projection on main; these three parameters keep projecting from their last authored value.

Microsimulation impact

We estimate no change in 2025, 2026 or 2032 on the default dataset, where none of the nine measured outputs moves. This branch adds DOR's 2026 limits, $26,200 and $39,200 on a joint return, and projects later years, where main keeps the 2025 limits of $25,400 and $38,100. Colorado's CollegeInvest subtraction totals $0 on main and on this branch in each year, so the higher limits change no record's subtraction or Colorado income tax. Congress keeps personal exemptions suspended, so exemptions total $0 on both and the corrected 2032 head-of-household phaseout threshold changes no record. The Biden capital gains proposal applies only when a user turns it on, so its threshold changes no record under current law. We compared main at bff5f062c5 with this branch at ec5403f3b4.

2025:

Output Main total Change Records changed (weighted)
co_collegeinvest_subtraction $0 $0 0 (0)
co_subtractions $16.69bn $0 0 (0)
co_taxable_income $275.2bn $0 0 (0)
co_income_tax $10.18bn $0 0 (0)
exemptions $0 $0 0 (0)
state_income_tax $548.1bn $0 0 (0)
income_tax $2,422.0bn $0 0 (0)
household_benefits $1,954.9bn $0 0 (0)
household_net_income $16,440.5bn $0 0 (0)

2026:

Output Main total Change Records changed (weighted)
co_collegeinvest_subtraction $0 $0 0 (0)
co_subtractions $17.70bn $0 0 (0)
co_taxable_income $298.4bn $0 0 (0)
co_income_tax $12.71bn $0 0 (0)
exemptions $0 $0 0 (0)
state_income_tax $576.6bn $0 0 (0)
income_tax $2,594.9bn $0 0 (0)
household_benefits $2,060.1bn $0 0 (0)
household_net_income $17,246.4bn $0 0 (0)

2032:

Output Main total Change Records changed (weighted)
co_collegeinvest_subtraction $0 $0 0 (0)
co_subtractions $23.04bn $0 0 (0)
co_taxable_income $375.6bn $0 0 (0)
co_income_tax $15.80bn $0 0 (0)
exemptions $0 $0 0 (0)
state_income_tax $725.7bn $0 0 (0)
income_tax $3,401.9bn $0 0 (0)
household_benefits $2,748.4bn $0 0 (0)
household_net_income $21,002.6bn $0 0 (0)

Axiom

axiom: TheAxiomFoundation/rulespec-us#2290 queued (C.R.S. 39-22-104(4)(i)(II)(B)-(C), Colorado's CollegeInvest limits; no corpus manifest holds DOR's November 2026 bulletin, the source of the 2026 limits, so the signed encoder can't run yet. The exemption threshold changes, to CBO's 2032 forecast and to projections after 2035, need no encoding: section 151(d)(5) sets the exemption amount to zero, and TheAxiomFoundation/rulespec-us#42 tracks 151(d)(3)'s reference to the removed 68(b) amount. The Biden threshold belongs to a Treasury proposal that PolicyEngine models as an optional reform.)

🤖 Generated with Claude Code

MaxGhenis and others added 10 commits November 12, 2026 00:28
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Keep both sets of personal exemption cases in exemptions.yaml: main's 21
for 2017 through 2026 and this branch's five for 2032 through 2038. The
merged file holds 26 cases.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
The exemptions formula on main counts a fraction of a phaseout step as a
full step, following IRC 151(d)(3)(B). A case that tests only one full
step above its threshold then passes for any threshold within that step.

Add adjusted gross income $1 above the threshold to each of the four
restored-exemption cases. The joint case again separates rounding the
2038 threshold down to $516,800 from rounding it to the nearest $50.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@MaxGhenis
MaxGhenis marked this pull request as ready for review November 15, 2026 23:20
@MaxGhenis MaxGhenis closed this Nov 16, 2026
@MaxGhenis MaxGhenis reopened this Nov 16, 2026
@MaxGhenis MaxGhenis closed this Nov 16, 2026
@MaxGhenis MaxGhenis reopened this Nov 16, 2026
@MaxGhenis
MaxGhenis merged commit 870bd46 into main Nov 16, 2026
77 of 81 checks passed
@MaxGhenis

Copy link
Copy Markdown
Contributor Author

US + core hub merge audit, us#10825 at ec5403f3b4c9138e933b34c81c8196f5f36f240c (squash): Activate uprating for filing-status parameter children.

  • Reviews:
    • The PR has no GitHub reviews. Three independent review rounds ran outside GitHub, with two fix rounds between them.
    • The first round, at 7d3b7c5ae6, read C.R.S. 39-22-104(4)(i)(II)(B)-(C), historical IRC 68(b), P.L. 119-21 and Treasury's FY2025 proposal, and checked the thresholds against CBO's January 2025 CSV. It ran 39 tests on the head but not the proposal's YAML file, which CI later ran. On the merge base 935b24ff4f, the head's tests failed 18 of 19 in the parameter file, 1 of 7 in the guard file, 2 of 3 exemption cases and 5 of 10 Colorado cases. Three households, for the Colorado subtraction, a restored exemption and the active proposal, matched its hand figures on both trees. The Colorado single filer, with $50,000 in 529 contributions, owed $35.20 less state income tax in 2026.
    • It couldn't verify the 2026 Colorado limits, since DOR's site answered with a bot check, and the description had no impact estimate. It also found that the 2032 head-of-household threshold read $421,050, where CBO's CSV gives $421,200. The preservation test sampled 20 of the 115 authored entries. A mutant that rounded joint thresholds to the nearest $50 passed the three exemption cases then in the PR, and the property test caught it. Justia links remained in two Colorado files.
    • The first fix round, through 70d9185227, cited DOR's November 2026 bulletin, corrected the 2032 value to $421,200, made the preservation test check all 115 authored entries, added a 2038 joint case and replaced the remaining Justia links.
    • The second round, at 70d9185227, confirmed each of those fixes in the diff, and the 2026 limits in DOR's bulletin. It recomputed the 2027 Colorado limits, the 2036 and 2038 exemption thresholds and the proposal's 2025 thresholds by hand, and checked the fix round's test logs. It asked for CI results under Core 3.36 and a check of the merge with main, and listed four corrections to the description. The second fix round made three, and the Claude Code footer it asked to delete stays.
    • The second fix round merged main at bff5f062c5. Git's one conflict, in exemptions.yaml, set main's 21 cases from Correct historical federal taxes and add annual 2017–2026 tests #10854 against this PR's five, and the merge keeps all 26. Since Correct historical federal taxes and add annual 2017–2026 tests #10854, the exemptions formula counts a fraction of a phaseout step as a full step, so the round added a point $1 above the threshold to cases 23 to 26. A check of the round found no required fix, and Ruff lint passed on the changed Python files.
    • The third round, at ec5403f3b4, required no correction. It confirmed that CI run 43774342022 passed all 37 jobs on the merge with bff5f062c5, with Core 3.36.1 and 180 focused cases. It confirmed that the merge keeps all 26 exemption cases under unique names, and recomputed each threshold and output in cases 23 to 26. A read-only merge analysis found no conflict with main at aead96b524.
    • Before merge, I made the description match the tested head. It now gives CI's per-file counts and shards, ties the Colorado tax cut to contributions above the 2025 limits, names the Department of Higher Education's and DOR's roles under (II)(C) and links a queued Axiom issue. It also gives the impact estimate that the first round found missing.
  • Methodology choices for maintainers:
    • Under (II)(C), the Colorado limits change each year with state college tuition, room and board costs, which the Department of Higher Education calculates, and DOR may round them to $100. The model anchors projections to DOR's 2026 limits and projects later years with the IRS inflation index in place of that cost change, rounded to the nearest $100. Each later DOR table should replace a projected year.
    • The exemption thresholds keep CBO's forecasts through 2035, with the 2032 head-of-household correction, and project later years from 2035 on the IRS index. Each child rounds down to $50, or $25 for separate filers, so separate filers keep half the joint threshold, as historical 68(b) requires. Section 151(d)(5) keeps the exemption amount at zero, so the thresholds change results in reforms that restore exemptions, not under current law.
    • Treasury's proposal indexes the threshold after 2024 but names no index or rounding interval. The model uses the IRS index with no rounding.
    • All three parameters keep chained projection from their last authored value and the IRS index's current measurement window, which get_irs_cpi averages August–July but documents (and IRS practice uses) the 12 months ending August #9105 and Index IRS parameters to the 12-months-ending-August chained CPI-U, with CBO's tax-year projections #9621 track. They don't use the base-year projection that Project statutory amounts from their base year with base-mode uprating #10783 added.
  • Gates checked live at merge:
    • gh pr checks exits 0;
    • the latest Pull request run at the head, 43944374859, passed and tested this head merged into main at 29f81ccc36;
    • MERGEABLE;
    • not a draft;
    • no reviewer's latest review requests changes;
    • the semantic-overlap guard: since 29f81ccc36, main has changed no area this PR touches. Exclude dependents' deducted health premiums from Schedule A #10831 changed three itemized medical expense variables and a test file under gov/irs/income/taxable_income before that base, and the latest Pull request run tested this head with them. Between the PR's merge base bff5f062c5 and main at 29f81ccc36, main changed 187 files and none of this PR's 11. Of the 21 parameter files main changed, 18 carry no breakdown and no uprating. Wyoming's new Medicaid parent income limit and Nebraska's refundable School Readiness credit amount carry a breakdown and no uprating. North Dakota's stillborn subtraction carries uprating and no breakdown, so none of the 21 gives the new guard test a parent to flag. Main's system.py change adds a marital-unit step and leaves parameter processing alone, and its version bumps change no policyengine-core version in the lock file. me_personal_exemption_deduction now returns federal exemptions for 2015 to 2017, years whose thresholds this PR leaves alone. git merge-tree writes a merge of the head into 29f81ccc36 with no conflict;
    • the merge pins the head.
  • Impact: we estimate no change in 2025, 2026 or 2032 on the default dataset, where none of the nine measured outputs moves. Colorado's CollegeInvest subtraction totals $0 on both sides in each year, so DOR's higher 2026 limits change no record, and the exemption and proposal thresholds reach no record under current law. Main bff5f062c5 against ec5403f3b4.
  • Next:
    • The Colorado subtraction applies one limit to a tax unit's total 529 contributions, where DOR applies it per beneficiary and counts ABLE contributions against the same limit. co_able_contribution_subtraction caps ABLE contributions on their own, and its cap stores no 2026 limit and still cites Justia.
    • start.yaml still cites IRC 151(d)(3)(B) through a law.cornell.edu link, which a govinfo link can replace.
    • Nobody has run, against the YAML cases at this head, the two mutants that the added points target. Rounding the joint threshold to the nearest $50 should fail case 26 at its third point, and restoring $421,050 for 2032 should fail case 24 at its first, second and fourth points.
  • Axiom: Encode Colorado's CollegeInvest contribution subtraction and its annually adjusted limits (C.R.S. 39-22-104(4)(i)(II)(B)-(C)) (PE-US #10825) TheAxiomFoundation/rulespec-us#2290 queued, a new pe-parity issue for C.R.S. 39-22-104(4)(i)(II)(B)-(C), with companion rows from this PR's Colorado YAML for 2022 and 2026. The signed encoder can't run yet: no corpus manifest holds DOR's November 2026 bulletin, the source of the 2026 limits, and the pinned release carries no DOR guidance or forms. The exemption threshold changes need no encoding, since section 151(d)(5) sets the exemption amount to zero, and Model 151(d)(3) phaseout after 68(b) applicable amount was removed TheAxiomFoundation/rulespec-us#42 tracks 151(d)(3)'s reference to the removed 68(b) amount. The proposal's threshold belongs to a Treasury proposal, not law.

🤖 Generated with Claude Code

@MaxGhenis
MaxGhenis deleted the fix-breakdown-parameter-uprating branch November 16, 2026 21:14
MaxGhenis added a commit that referenced this pull request Nov 17, 2026
No commit on main since the last merge changes a file this branch
changes, so the merge has no conflict.

Main adds tests in #10825 that read projected values of
gov.irs.uprating: the FY2025 capital gains threshold in 2025,
Colorado's CollegeInvest limits in 2027 and the legacy personal
exemption thresholds after 2035. Their expectations follow main's
August-to-July index. The next commit moves them to this branch's
September-to-August index.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MaxGhenis added a commit that referenced this pull request Nov 17, 2026
The 2025 and 2024 index values average 12 months of chained CPI-U that
sum to 2,077.121 and 2,021.359. The FY2025 capital gains threshold
becomes 1,000,000 * 2,077.121 / 2,021.359 = $1,027,586.3911358645, and
$513,793.1955679323 for separate filers. Move the two YAML cases'
inputs to less than $1 below the threshold, the threshold itself and
less than $2 above it, and update the expected tax and the Python test to match.

Colorado's 2027 limits stay at $27,000 and $40,400: 26,200 * 2,190.008 /
2,125.365 = 26,996.8733 and 39,200 * the same ratio = 40,392.2684.
The 2036 exemption thresholds keep their values under the factor
217.68 / 213.471. Update the arithmetic in both comments.

The 2038 joint case no longer separates rounding down from rounding to
the nearest $50: 487,400 * (217.68 / 213.471)^3 = 516,802.32 rounds to
$516,800 either way. Move it to 2039, where 487,400 * (217.68 /
213.471)^4 = 526,992.10 rounds down to $526,950 and to the nearest $50
at $527,000. AGI of $526,951 then loses one 2% step.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant