[MaxGhenis]

What and why

From January 2025, tx_ccs_payment_rate reads Texas Child Care Services rates from two eight-age-group tables: rates_expanded_age_groups/bcy2025.csv until September 2025 and bcy2026.csv after that. This PR:

• corrects 127 cells in bcy2026.csv and one in bcy2025.csv to match the Texas Workforce Commission's (TWC) tables;
• fills the relative care and blended rates that TWC's January to September 2025 table doesn't print, for which main pays $0;
• adds invariant tests over all 8,512 cells of each table.

Corrected cells

BCY 2026

TWC's table (WD Letter 08-25, Attachment 1, effective October 1, 2025) differs from the CSV in 127 cells across 13 Boards. This PR raises 111 of them and lowers 16, by a median of $0.38 a day.

Board	Cells
Permian Basin (LCCH 56, LCCC 7)	63
South Texas	15
Deep East, Rural Capital	11 each
Middle Rio Grande, Northeast Texas	5 each
North Texas, South Plains	4 each
Golden Crescent	3
Gulf Coast, Southeast	2 each
Lower Rio Grande Valley, North Central	1 each

This PR raises Permian Basin LCCC Four-Star, 12-17 months full time, from $39.22 to $42.80, its largest increase. It lowers Rural Capital LCCH regular, 4-year-old blended, from $39.42 to $39.39. At $39.42, that cell's Two-Star and Texas School Ready rates fell short of the 5 percent markup in 40 TAC §809.20(c).

BCY 2025

TWC's eight-age table prints $22.80 for Golden Crescent LCCH regular, 6-13 years part time, where the CSV held $22.20.

Extraction

pdftotext -layout reads each table by column order, and pdfplumber assigns each rate to its printed header by position. They agree on all 8,512 BCY 2026 cells and all 6,720 printed BCY 2025 cells, and the CSVs now match them. TWC's Word copy of Attachment 1 matches all 8,512 BCY 2026 cells too.

Editions

TWC's Word attachment (saved September 30, 2025), TWC's PDF (created October 6, 2025) and the childcare.texas.gov copy (saved March 9, 2026) print the same BCY 2026 rates. South Texas's one-page sheet (created September 30, 2025) prints the same South Texas rates. South Plains's provider page prints TWC's values in three of its four corrected cells and has no Texas School Ready row for the fourth. None matches the old values. South Plains's page also prints $36.36 for LCCH Four-Star, 3 years full time, where TWC prints $36.30, and we keep TWC's value.

bcy2026.csv kept the same bytes from its first appearance, as rates_v2/bcy26.csv in 65a83a13ce (February 11, 2026), until this PR. f3343c7071 and 8180760814 only moved and renamed it. We couldn't establish where its values came from, and they don't form one coherent table:

• 43 of the 127 fall below TWC's BCY 2025 rate. Under WD 08-25, TWC kept the prior year's rate wherever a new one would fall, and its table has no such cell.
• 42 blended cells in the old table miss WD 02-20's formula on its own part-time and full-time rates, and this PR changes each one or its inputs. TWC's table has none.
• 50 of Permian Basin's 63 carry Heart of Texas's value for the same provider, rating and column.

That pattern fits transcription or extraction error, though an unarchived edition stays possible.

January to September 2025

_lookup_expanded_rate returns $0 where a table has no cell. TWC's eight-age table prints no relative rows and no blended columns, so main pays $0 for relative care and blended schedules in those nine months and pays again from October 2025. Main's tx_ccs_payment_rate.yaml expects that $0 for a Capital Area relative provider in January 2025. bcy2025.csv now holds 8,512 cells, with the same rows, columns and order as bcy2026.csv.

Relative care

Under 40 TAC §809.20(a)(1)(D), each Board must set full-day and part-day rates for relative providers, and in WD 19-24, Change 1, TWC leaves those rates to Boards. So each Board had relative rates in these months that TWC's table doesn't list. This PR fills the 448 relative full-time and part-time cells (28 Boards × 8 ages × 2 schedules) with each Board's rates from TWC's four-age BCY 2025 table. We map the letter's four age groups onto the eight: Infant 0-17 months, Toddler 18 months to 2 years, Preschool 3-5 years and School Age 6-13 years.

In the same letter, TWC ends the four-age rates on January 12, 2025, and we carry their relative rates forward to September 30, 2025. TWC's BCY 2026 table matches the carried rates in 447 of 448 cells. The exception, Brazos Valley school-age part time, moves from $5.35 to $5.23. North Texas's Board sheet effective 1-13-2025 prints relative rates for all eight ages, and all 16 match, as do its 210 regulated cells.

Blended schedules

TWC defines the blended-day unit in 40 TAC §809.93(f)(3) and sets the formula in WD Letter 02-20: (175 × part-time rate + 30 × full-time rate) / 205. In its 2026 rule adoption, the Commission says the blended-rate method its rules describe "has been in place since 2010". We apply WD 02-20's formula, rounding half up to the cent, to ages 4, 5 and 6-13, the ages BCY 2026 prints, in all 448 rows (28 Boards × 16 provider type and rating rows). It reproduces all 448 school-age blended rates in the four-age table and all 1,344 blended rates in BCY 2026. In its provider handbook, revised July 3, 2025, Alamo points providers to TWC's eight-age table for rates from January 13, 2025 and computes blended rates with the same formula.

Relative labels

TWC labels the relative row "None" for 27 Boards and "Reg" for Alamo in its four-age table, and "Reg" for all 28 in BCY 2026. tx_ccs_payment_rate now reads that rate under either label, so it pays a relative provider entered as REG before 2025 and one entered as NONE from 2025. Main pays each of them $0. The model still pays $0 for relative care at a star rating.

Impact

We ran one household for each kind of change: a parent earning $25,000 with one child in care 20 days a month and $20,000 a year in pre-subsidy care costs. The household qualifies for the subsidy in every scenario. Each row compares this branch with main at 4e71481e19.

Scenario	Monthly rate from January	Texas subsidy, year	SNAP, year	SPM net income, year	Change in SPM net income on #10569
Relative infant, Capital Area, 2025	$0 → $288.80	$597 → $2,389	$6,462 → $6,462	$17,476 → $17,476	+$1,792
Relative 8-year-old part time, Brazos Valley, 2025	$0 → $107.00	$45 → $200	$6,462 → $6,462	$16,392 → $16,392	+$156
Blended 8-year-old, regular center, Capital Area, 2025	$0 → $874.40	$2,446 → $9,509	$6,462 → $4,344	$16,392 → $14,274	+$4,944
Blended 4-year-old, regular center, Capital Area, 2025	$0 → $916.80	$2,553 → $9,996	$6,429 → $4,197	$15,993 → $13,761	+$5,212
Part-time 8-year-old, regular home, Golden Crescent, 2025	$444 → $456	$4,359 → $4,467	$5,889 → $5,856	$15,819 → $15,786	+$75
Full-time 15-month-old, Four-Star center, Permian Basin, 2026	$784.40 → $856	$8,364 → $9,223	$4,845 → $4,590	$14,081 → $13,826	+$604
Full-time 3-year-old, regular home, Permian Basin, 2026	$620 → $636	$6,391 → $6,583	$5,433 → $5,382	$14,669 → $14,618	+$141
Blended 4-year-old, regular home, Rural Capital, 2026	$788.40 → $787.80	$8,412 → $8,405	$4,830 → $4,833	$14,066 → $14,069	−$4

SPM net income falls as the subsidy rises in five of the eight scenarios, and SNAP accounts for the whole change. snap_dependent_care_deduction reads childcare_expenses, which nets out the subsidy, so the model lowers the deduction, and SNAP with it, as the subsidy rises. On main, spm_unit_benefits doesn't count tx_child_care_subsidies, and spm_unit_capped_work_childcare_expenses reads pre-subsidy spending, so nothing offsets the SNAP loss. household_net_income moves the same way, because household_state_benefits doesn't list the Texas subsidy either. #10569 lowers the SPM expense by the subsidy, and #9503 adds the subsidies to household benefits; this PR changes neither. The last column shows this PR's effect with #10569's head (652e968a94) merged.

Population

The default dataset (populace_us_2024.h5) has no childcare_attending_days_per_month column, and tx_ccs_payment_rate multiplies the daily rate by those days. For a household that omits the input, the model computes a $0 rate and $0 tx_ccs on this branch. So the model pays $0 in Texas CCS on every record of that dataset, with or without this PR.

Invariants

test_tx_ccs_expanded_rates_table.py states these and checks them across every cell of both CSVs:

• Completeness: 28 Boards × 16 provider type and rating rows × 19 age and schedule columns give 8,512 cells, and both years list the same rows in the same order.
• Texas Rising Star markups (40 TAC §809.20(c)): Two-Star, Texas School Ready, Three-Star and Four-Star rates exceed the regular rate by at least 5, 5, 7 and 9 percent. Our rounding leaves six BCY 2025 blended cells for ages 4 and 5 under a cent short, and no TWC table prints those rates. A test pins their count and bounds their size.
• Star gaps (40 TAC §809.20(d)): each star level exceeds the one below by at least 2 percent of the regular rate.
• Texas School Ready: those rates equal Two-Star rates.
• Blended formula (WD 02-20): every blended cell equals (175 × part time + 30 × full time) / 205, rounded half up.
• Schedule ordering: part time ≤ full time, and part time ≤ blended ≤ full time. No rule requires this, but both tables satisfy it, so the test catches a value in the wrong column.
• No cuts (WD 08-25): every regulated BCY 2026 rate meets or exceeds its BCY 2025 rate.
• Age crosswalk (WD 19-24, Change 1): each four-age regulated rate equals the lowest eight-age rate in its group.
• Four-age agreement: BCY 2025 relative rates and 6-13 blended rates equal the four-age rates in rates.yaml.
• Relative rates across years: only Brazos Valley's school-age part-time and blended rates change from BCY 2025 to BCY 2026.
• Lookup round trip: get_rates returns every CSV cell, and a Hypothesis test checks that _lookup_expanded_rate returns each cell in input order and $0 for combinations the tables don't print.

These checks catch 124 of the 127 old BCY 2026 values and the Golden Crescent value when we insert each one alone into the corrected table. The two-parser comparison with TWC's tables supplies the rest, because a wrong value can satisfy every rule.

Limits

• We searched all 28 Boards' sites for sheets effective from January 13 to September 30, 2025, and found one that prints relative rates, North Texas's, and none that prints blended amounts. We couldn't open several Boards' files, so other sheets may exist. A search result describes a Brazos Valley eight-age sheet headed December 9, 2024, with a relative row between $5.35 and $12.76, the range of that Board's four-age row. We couldn't retrieve it, and every TWC copy of the eight-age table we hold (sha256 a4513140…) has no relative rows.
• TWC extends the blended-day unit to children in pre-K, Head Start and Early Head Start in §809.93(f)(3). In WD 02-20, it computes blended rates for school-age children and children in full-day pre-K, and says three-year-olds in half-day pre-K typically need full-day care. Alamo, in its handbook, offers blended care to infants and toddlers in Early Head Start. TWC prints blended rates only from age 4 in BCY 2026 and only for school age in the four-age table. So the model pays $0 on a blended schedule for a child younger than 4 from 2025, or younger than 6 before 2025.
• TWC published BCY 2027 rates in WD Letter 16-26, effective October 1, 2026. This PR doesn't add them, so get_rates applies BCY 2026 rates from October 2026.

Tests

• 18 cases in tx_ccs_payment_rate.yaml replace the one that pinned $0 for BCY 2025 relative care. They pin relative and blended rates across September and October 2025, relative labels and star ratings before and from 2025, and corrected Golden Crescent, Permian Basin and Rural Capital cells. Their expected values come from TWC's tables and North Texas's sheet.
• UNCORRECTED_FOUR_AGE in the new test file holds the four Dallas County cells Add Texas CCS Board Contract Year 2024 provider payment rates #10598 corrects, and the crosswalk test skips a cell only while it holds the uncorrected value. TWC's four corrected values satisfy the crosswalk, so whoever lands the second of the two PRs removes that block. This head, Add Texas CCS Board Contract Year 2024 provider payment rates #10598's head and Add Texas CCS provider payment rates for October 2016 to September 2023 #10622's head merge without conflict.

axiom: TheAxiomFoundation/rulespec-us#2013 queued

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (b0a8c09) to head (02f4163).
⚠️ Report is 151 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main    #10636   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            1         1           
  Lines           15        36   +21     
  Branches         0         1    +1     
=========================================
+ Hits            15        36   +21     

Flag	Coverage Δ
unittests	100.00% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[MaxGhenis]

Ready for hub review
VERDICT: APPROVE_WITH_NITS

Head reviewed: 311f7b04e8add4fa79a00c9bcea4bcc00cd47937, against merge base 4e71481e19. The merge with current main (git merge-tree --write-tree origin/main 311f7b04e8) is clean, and main has not touched any Texas CCS file since the base.

Scope limit: read this first

This verdict covers the code, the tests, internal consistency, the impact runs and the merge with #10598. It does not cover whether the CSVs match TWC's printed tables. This review environment's network policy refused every source host (web.archive.org, twc.texas.gov, childcare.texas.gov, ntxworksolutions.org, southtexasworkforce.org, law.cornell.edu): each returned a 403 on CONNECT. I fetched no primary source, hashed none and extracted none. So I did not run the core data check: 8,512 BCY 2026 cells and 6,720 printed BCY 2025 cells against TWC's PDF and Word copy, the four-age relative rows, the North Texas sheet or the South Texas sheet. Every claim in the PR that depends on reading a source is marked unverifiable below. Someone who can reach those files should run that comparison before merge. Nothing I found suggests it will fail.

What I ran

• policyengine-core test policyengine_us/tests/policy/baseline/gov/states/tx/twc/ccs -c policyengine_us: 147 passed.
• pytest -q .../payment/test_tx_ccs_expanded_rates_table.py: 23 passed.
• policyengine_us/tests/code_health/test_source_conflicts.py: 103 passed.
• CI on this head: 43 check runs, 36 succeeded, 7 skipped (the umbrella and parent jobs), 0 failed. Passing runs include all baseline and contrib shards, Python shards 1-7, Microsimulation, Household API Partners and codecov patch and project.
• The 8 impact scenarios from the spec, on this head and on 4e71481e19.
• 17 label and schedule edge cases on both sides, plus one household with three children of mixed types (vectorized).
• The new test files copied onto main.
• The invariant tests with each old value put back alone, plus 785 seeded errors of my own.
• The merged tree of this head and Add Texas CCS Board Contract Year 2024 provider payment rates #10598's head 1ca5db7e38.

Findings

No P1 or P2.

P3. changelog.d/tx-ccs-expanded-rates.fixed.md:1 leaves out the relative-label change. The fragment covers the cell corrections and the January to September 2025 fill. It doesn't mention that a relative provider entered as REG now pays from October to December 2024, and one entered as NONE now pays from October 2025. I confirmed both changes on the base and the head (details below). Fix: add a clause such as "and pay relative care entered under either the Reg or None rating".

P3. The PR body's "42 blended cells" bullet reads as a subset of the 127, but it isn't one. The bullet sits under "they don't form one coherent table", right after "43 of the 127". The whole old bcy2026.csv has 42 blended cells that miss WD 02-20's formula on its own part-time and full-time rates. All 42 touch a corrected cell. But only 27 of those blended cells are themselves among the 127. In the other 15, the part-time or full-time input was the wrong value. Coincidentally, 42 blended cells are among the 127 too. Fix: say "the old table has 42 blended cells that miss…".

P3. PR body, file history. The body says "8180760814 only renamed it". That's true, but the file also moved between commits, from rates_v2/bcy26.csv (where 65a83a13ce added it) to rates_expanded_age_groups/bcy26.csv, before 8180760814 renamed it to bcy2026.csv. The bytes never changed: the blob is a360146c at all three points, so "kept the same bytes" holds. The committer date of 65a83a13ce is 2026-02-12T00:24Z (February 11 in US time zones). No change needed beyond wording.

P3 (gap main also had; the PR states it). No TWC table prints BCY 2025 blended rates for ages 4 and 5. The PR derives them with the WD 02-20 formula from January 2025, while preschool blended care before 2025 still pays $0. Six of the derived cells fall short of the 5 percent markup by less than a cent. The PR states this choice plainly: the __init__.py header says "for the ages BCY26 prints", the body's Limits section covers it, and test_derived_blended_shortfalls_stay_as_counted pins the six cells. I'd keep it. Paying a derived rate is closer to §809.93(f)(3) than paying $0, but it rests on reading the formula forward, not on a printed rate.

Checked and found correct

Old-value claims, recomputed from git show 4e71481e19:…/bcy2026.csv:

• 127 changed cells.
• Per-Board counts match the body's table exactly: Permian Basin 63 (LCCH 56, LCCC 7), South Texas 15, Deep East 11, Rural Capital 11, Middle Rio 5, Northeast Texas 5, North Texas 4, South Plains 4, Golden Crescent 3, Gulf Coast 2, Southeast 2, Lower Rio 1, North Central 1.
• 111 rise and 16 fall. The median absolute change is $0.38.
• The largest change is Permian Basin LCCC TRS4 age_12_17m_ft, $39.22 to $42.80.
• 43 of the 127 old values fall below the BCY 2025 rate.
• 50 of Permian Basin's 63 old values equal Heart of Texas's value in the same row and column.
• Rural Capital LCCH REG age_4yr_bt goes from $39.42 to $39.39. TRS2 and TSR are $41.37, which is 4.95 percent above the old value and 5.03 percent above the new one, so the body's markup statement holds.

BCY 2025 changes:

• Exactly one printed cell changes: Golden Crescent LCCH REG age_6_13yr_pt, $22.20 to $22.80.
• 1,792 cells were empty or missing on the base and are now filled: 448 relative ft/pt cells, 84 relative blended cells and 1,260 regulated blended cells. That is the body's 448 + 1,344.

Relative fill:

• I recomputed all 448 cells from rates.yaml RELATIVE/NONE with the age map 0-17 months Infant, 18 months to 2 years Toddler, 3-5 Preschool, 6-13 School Age: 448 of 448 match.
• 447 of 448 equal bcy2026.csv. The exception is Brazos Valley age_6_13yr_pt, $5.35 to $5.23.
• bcy2026.csv's relative rows are unchanged by this PR.
• I find the carry-forward the best reading available without a printed eight-age relative row. Its strongest support is internal: 447 of 448 cells match TWC's next table. I couldn't read the text of §809.20(a)(1)(D) or WD 19-24 Change 1, or check the North Texas 16 and 210 counts.

Blended fill:

• (175 × pt + 30 × ft) / 205, rounded half up, reproduces 1,344 of 1,344 bcy2026.csv blended cells and 1,344 of 1,344 bcy2025.csv blended cells.
• It also reproduces all 448 four-age school-age blended rates in rates.yaml (bcy2025.csv age_6_13yr_bt equals them, 448 of 448).
• On the base, the old bcy2026.csv has 42 misses.

Code (tx_ccs_payment_rate.py):

• The label normalization touches only RELATIVE with REG or NONE. Those map to NONE in the four-age table and REG in the CSVs. The October switch (bcy_year = year + (month >= 10)) is unchanged.
• _lookup_expanded_rate now takes a string rating. Its only callers are the formula and the new test, per grep.
• No input pays that shouldn't. Each of these gives $0 on both the base and the head:
  • a star-rated relative provider (TRS2 in Nov 2024 and Jun 2025, TRS4 in Oct 2025, TSR in Jun 2025);
  • an LCCC provider rated NONE (Nov 2024, Jun 2025, Oct 2025);
  • LCCC REG blended for age 3 (Jun 2025) and for age 4 (Nov 2024).
• No input stops paying.

Claim 7, Capital Area relative infant, full time:

Input	Base	Head
REG, Nov 2024	$0 (no REG key in the four-age tree, which resolves to 0)	$288.80
NONE, Nov 2024	$288.80	$288.80
NONE, Jun 2025	$0 (no relative row)	$288.80
REG, Jun 2025	$0	$288.80
NONE, Oct 2025	$0 (no NONE row; the merge's fillna(0) gives 0)	$288.80
REG, Oct 2025	$288.80	$288.80

• Vectorized: one household with a relative REG infant, an LCCC TRS4 infant and a relative NONE 3-year-old gives the right rate for each child in Nov 2024, Jun 2025 and Oct 2025.

Tests:

• All 15 new YAML expectations I recomputed equal CSV rate × 20, and the 16th, relative TSR, expects $0, which the head returns. That's 16 cases (one replaced, 15 added).

• On main, 14 of the 16 new cases fail. The two that pass are the $0 cases: relative TSR and LCCC 3-year-old blended. Each failure is a change this PR makes: the relative fill, the labels, the blended fill, Golden Crescent, Permian Basin or Rural Capital.

• The new pytest file errors at collection on main, because Decimal('') meets main's empty blended cells.

• With the head's bcy2025.csv swapped into main, four tests fail: the 2026 markups, the 2026 blended formula, no-cuts, and the lookup (a signature change).

• Each invariant reads every cell, and none can pass vacuously. The crosswalk test asserts checked >= 3,356. UNCORRECTED_FOUR_AGE skips a cell only while it holds the listed value. The 2025 ages-4-and-5 allowance is bounded by the exact-count test, so a seventh shortfall fails it.

• With each old value put back alone, 124 of 127 are caught, and so is Golden Crescent's $22.20. The three that pass are Lower Rio LCCC REG 12_17m_pt 35.20, North Central LCCC TRS3 12_17m_ft 59.98 and Rural Capital LCCC TRS4 18_23m_pt 48.53. Only a source comparison catches those.

• My own seeded errors (seed 7, 200 cells each) were caught at these rates:

  • +1 percent: 188 of 200;
  • −1 percent: 194 of 200;
  • +5 cents: 176 of 200;
  • swap with the next column: 185 of 185.

  These are in line with the body's 369/400, 389/400, 338/400 and 373/387.

Impact: all 8 scenarios reproduce the body's table to the dollar on both sides: monthly rate, subsidy, SNAP and SPM net income. I skipped the #10569 column.

Mechanism: the SNAP and SPM explanation holds.

• snap_dependent_care_deduction adds childcare_expenses, which is pre-subsidy spending minus child_care_subsidies, and gov.hhs.ccdf.child_care_subsidy_programs lists tx_child_care_subsidies.
• spm_unit_benefits doesn't list tx_child_care_subsidies.
• spm_unit_capped_work_childcare_expenses reads spm_unit_pre_subsidy_childcare_expenses.
• gov.household.household_state_benefits has no Texas child care entry.
• In every row, the drop in SPM net income equals the drop in SNAP.

Population: childcare_attending_days_per_month is an int with no formula and no default, so it is 0. The rate is daily rate × days, and tx_ccs is max(min(expense, rate) − copay, 0), which is $0 when the input is missing. I could not list the dataset's columns, because the dataset host is blocked too, so the claim that the column is absent is unverifiable.

Repo rules:

• The fragment is top-level, with type fixed.
• References: one entry per document, with #page=N only in the href.
• No partner-test files changed.
• The source-conflict test passes.
• The axiom: line has the <rulespec issue> queued form. I can't see the rulespec-us issue from here, so I didn't check whether it is dispatch-ready.

Fit with #10598:

• git merge-tree gives the same clean tree in both orders (574215e1).
• In that tree, both rate-table test files pass (31 tests), and so does the Texas CCS YAML folder (170 cases).
• The four Dallas County cells hold Add Texas CCS Board Contract Year 2024 provider payment rates #10598's values (46.60, 39.43, 34.80, 34.00), so the crosswalk test checks them instead of skipping them. UNCORRECTED_FOUR_AGE becomes dead code, which the body says the second PR to land should remove.
• The reference tuple merges to five entries without conflict.

Unverifiable here (source hosts blocked)

• Every source hash.
• 8,512 BCY 2026 cells against TWC's PDF, Word copy and childcare.texas.gov copy.
• 6,720 BCY 2025 printed cells against the eight-age PDF.
• The South Texas sheet against bcy2026.csv.
• North Texas's 16 relative and 210 regulated cells.
• That the eight-age table prints no relative rows or blended columns.
• That TWC labels relative rows "Reg" or "None" "by table and by Board". In the repo, all 28 four-age Boards use NONE and all CSV Boards use REG.
• The quoted text of WD 19-24 Change 1, WD 02-20, WD 08-25, §809.20 and §809.93, and the 2026 adoption's "has been in place since 2010".
• Edition dates and the Brazos Valley search-result description.
• The dataset's columns.
• The Count state child care subsidies in the SPM once #10569 column of the impact table.

---

Generated by Claude Code

[MaxGhenis]

We merged this at 02f4163256 with admin rights, because GitHub blocks merging without an approving review. Reviews and checks at merge:

• An Opus 5.5 review on Subfleet requested changes at 311f7b04e8, and a GPT-6.1 Sol review on Subfleet checked each fix at 02f4163256.
• The Sol review approved 02f4163256 with four nits, all on the PR description, and we applied them.
• A cloud review approved 311f7b04e8 with nits on code, tests and impact (comment above).
• All 36 checks pass at 02f4163256. The head merges without conflict with main (a549fbce44), Add Texas CCS Board Contract Year 2024 provider payment rates #10598 (b08c480125) and Add Texas CCS provider payment rates for October 2016 to September 2023 #10622 (8e7782eb3c), and both rate-table test files pass on the combined tree.

Whichever of #10598 and #10622 lands first brings TWC's Dallas County values to main and can drop UNCORRECTED_FOUR_AGE from test_tx_ccs_expanded_rates_table.py.