New Hampshire 2026 SGP Results

NH SAS (Grades 3–8) and WIDA ACCESS (Grades K–12)

Damian W. Betebenner

National Center for the Improvement of Educational Assessment

NH TAC Meeting · October 19, 2026

AI Contribution & Authorship Statement

These analyses and this presentation represent the work of the author with substantive and continuous assistance from AI-based systems (Anthropic’s Claude), integrated natively into the analysis, revision, and preparation of this deck.

The use of AI here is not incidental or peripheral — it is essential, embedded, and reflective of a methodological shift toward AI-native scholarship. This statement is offered in the spirit of methodological clarity and epistemic honesty.

  • Verification of inputs: Every figure and factual claim is generated from cached aggregates of the student-level files. Claims that cannot be verified from those files are deliberately not asserted.

  • Computational modules: AI contributed to the R summary pipeline, the ggplot2 figures, and the verification of reported numbers against the analysis logs.

  • Authorship & responsibility: The author retains full responsibility for all claims, analyses, and recommendations (ICMJE/AERA norms). AI is not listed as an author but its instrumental role is explicitly acknowledged.

Roadmap

  1. What was run — NH SAS cohort and baseline analyses
  2. 2026 results — overall growth and model fit
  3. Growth in context — pandemic recovery and the 2018–19 baseline
  4. Growth equity — results by student group
  5. Target attainment — catch-up / keep-up
  6. WIDA ACCESS — growth to the Level 4.5 exit standard
  7. Questions for the TAC

110,854 students received a 2026 NH SAS growth percentile (grades 4–8, ELA and mathematics). 4,404 English learners received a 2026 WIDA baseline growth percentile. NH SAS results use the student-level file received July 9, 2026.

SGP results for 2026

What was run: NH SAS (grades 4–8)

Two SGP analyses were conducted on the 2026 NH SAS data (ELA and Mathematics, grades 3–8; growth from grade 4):

  • Cohort-referenced — growth norms built from the most recent cohorts; by construction these center mean/median SGP on 50.
  • Baseline-referenced — historical coefficient matrices from the pre-pandemic 2018–19 administration (with 2017–18 priors). These are not centered on 50 and let us examine COVID-era academic recovery, changes in rates of learning, and possible scale drift.
  • Projections (lagged and straight, three-year catch-up / keep-up and move-up / stay-up) were calculated using both norm sets.

Science and grade 11 are on the file but are not included in the SGP analyses.

Cohort norms answer “how did this student grow relative to academic peers this year?” Baseline norms answer “how does that growth compare to pre-pandemic rates of learning?”

Provenance & reproducibility

Reproducible pipeline

This deck is generated from cached summaries of the student-level LONG data. When NHDOE supplies revised data, re-running the pipeline refreshes every number and table automatically.

2026 SGP results at a glance

Grade ELA Mathematics
Cohort Baseline Cohort Baseline
4 49.9/50/11,153 52.2/53/11,153 49.9/50/11,225 51.3/51/11,225
5 50.0/50/11,092 54.7/56/11,092 50.0/50/11,148 49.3/49/11,148
6 49.8/50/11,045 51.7/52/11,045 49.8/50/11,122 51.4/52/11,122
7 50.0/50/11,106 57.2/61/11,106 49.9/50/11,210 54.1/56/11,210
8 49.9/50/10,825 54.2/56/10,825 49.8/50/10,928 54.9/57/10,928
Each cell: mean / median / N of student growth percentiles, 2026. Cohort norms center on 50 by construction; baseline norms are referenced to 2018–19 rates of learning. Data received July 9, 2026.

Reading the 2026 baseline results

Above pre-pandemic rates (baseline mean > 52)

  • ELA Grade 7 (mean 57.2)
  • Mathematics Grade 8 (mean 54.9)
  • ELA Grade 5 (mean 54.7)
  • ELA Grade 8 (mean 54.2)
  • Mathematics Grade 7 (mean 54.1)
  • ELA Grade 4 (mean 52.2)

Below pre-pandemic rates (baseline mean < 48)

  • None materially different

Cohort-referenced results are, as expected, centered on 50 with even conditional coverage. The baseline results are the diagnostic lens — they say whether 2026 growth matches the 2018–19 rate of learning.

Growth distribution: Low / Typical / High

Stacked bar chart of the share of students at low, typical and high growth by grade, in panels for each subject and for cohort and baseline norms. Dashed lines mark one-third and two-thirds; cohort norms split about evenly, while baseline norms shift with recovery or drift.

Model fit and conditional coverage

What these plots do and do not diagnose

Each plot shows, for every decile of prior achievement, the share of students falling in each growth-percentile decile. Two different things move those cells, and only one of them is a fitting problem.

What they diagnose

Systematic tilt across prior achievement. If growth percentiles drift upward or downward as prior achievement rises, the conditional model is not covering the range of prior achievement evenly. That is misfit, and it is actionable.

What they cannot diagnose

Departure from exact uniformity. Scores are discrete. Students sharing a prior and a current score must receive the same growth percentile, so the conditional distribution cannot be uniform and the cells cannot all sit at 10%. That lumpiness follows from fitting a continuous quantile model to a discrete scale.

The diagnostic content is the tilt, not the distance from 10%. For baseline norms a level shift is neither misfit nor uniformity: it is the change in the prior-to-current relationship that the comparison exists to detect.

NH SAS ELA — cohort model fit, 2026

ELA Grade 4
ELA Grade 4
ELA Grade 5
ELA Grade 5
ELA Grade 6
ELA Grade 6
ELA Grade 7
ELA Grade 7
ELA Grade 8
ELA Grade 8

NH SAS Mathematics — cohort model fit, 2026

Mathematics Grade 4
Mathematics Grade 4
Mathematics Grade 5
Mathematics Grade 5
Mathematics Grade 6
Mathematics Grade 6
Mathematics Grade 7
Mathematics Grade 7
Mathematics Grade 8
Mathematics Grade 8

Cohort vs. baseline fit — ELA Grade 7

ELA Grade 7 - Cohort norms
ELA Grade 7 — Cohort norms
ELA Grade 7 - Baseline norms
ELA Grade 7 — Baseline norms

Growth in context

Where were they during the pandemic?

The 2026 tested grades sat in very different places when schooling was disrupted. There is no 2019_2020 NH SAS administration.

Grade in 2026 Spring 2020 2020–21 (first post-COVID test)
Grade 8 Grade 2 Grade 3
Grade 7 Grade 1 Grade 2
Grade 6 Kindergarten Grade 1
Grade 5 Pre-K Kindergarten
Grade 4 Not yet in school Pre-K
Grade 8 in 2026 spent early-elementary years in disrupted schooling; grade 4 had not yet entered the tested progression.

Trend: same grade over time (baseline SGP)

Line chart of mean baseline SGP by year, 2022 to 2026, one panel per grade with a line for each subject. A dashed line at 50 marks the 2018–19 rate of learning.

ELA Grade 7 over time — goodness of fit

ELA Grade 7 - 2023
ELA Grade 7 — 2023
ELA Grade 7 - 2024
ELA Grade 7 — 2024
ELA Grade 7 - 2025
ELA Grade 7 — 2025
ELA Grade 7 - 2026
ELA Grade 7 — 2026

Growth equity

Growth by student group (NH SAS, 2026)

Group ELA Mathematics
Cohort Baseline Cohort Baseline
English Learner
Non-English Learners (Non-EL) 50.0/50/28.9/52,303 54.1/56/28.3/52,303 49.9/50/28.9/52,519 52.2/53/29.0/52,519
English Learners (EL) 49.1/48/28.6/2,918 51.8/53/28.1/2,918 49.5/50/28.9/3,114 51.1/51/28.8/3,114
Students w/ Disability
Students without Disabilities (Non-SWD) 51.4/52/28.7/44,185 55.9/58/28.0/44,185 50.9/51/28.7/44,487 53.4/55/28.8/44,487
Students with Disabilities (SWD) 44.1/42/28.7/11,036 46.4/45/28.2/11,036 45.8/44/29.3/11,146 47.2/46/29.2/11,146
Economically Disadv.
Economically Disadvantaged 46.9/46/29.0/13,673 50.0/50/28.5/13,673 46.4/45/29.0/13,886 48.1/47/29.1/13,886
Not Economically Disadvantaged 50.9/51/28.8/41,548 55.3/57/28.1/41,548 51.0/52/28.8/41,747 53.5/55/28.8/41,747
mean / median / SD / N, NH SAS grades 4–8 pooled, 2026. Cells with N < 10 are suppressed. For reference, SGPs uniformly distributed over the integers 1–99 have SD = √((99² − 1) / 12) ≈ 28.6.

Growth by gender and race/ethnicity (NH SAS, 2026)

Group ELA Mathematics
Cohort Baseline Cohort Baseline
Gender
Male 48.5/48/29.2/28,330 52.3/53/28.7/28,330 49.9/50/29.2/28,562 52.2/53/29.3/28,562
Female 51.4/52/28.5/26,815 55.7/57/27.7/26,815 49.9/50/28.5/26,994 52.1/53/28.6/26,994
Race/Ethnicity
White 49.9/50/28.8/45,358 54.0/55/28.2/45,358 50.0/50/28.9/45,567 52.3/53/29.0/45,567
Asian 55.3/57/28.6/1,566 59.8/63/27.6/1,566 56.0/58/28.2/1,610 58.9/62/28.0/1,610
Black or African American 48.0/48/29.1/1,080 51.0/52/28.7/1,080 47.3/47/28.0/1,111 48.7/49/28.0/1,111
Hispanic or Latino 48.6/48/29.2/4,824 51.7/53/28.7/4,824 47.2/46/29.0/4,943 49.0/48/29.0/4,943
Two or More Races 50.5/51/28.6/2,191 54.9/57/27.8/2,191 50.2/51/29.0/2,199 52.6/53/29.1/2,199
American Indian or Alaska Native 51.2/52/28.8/94 55.1/56/28.4/94 55.1/58/29.5/93 57.5/62/29.3/93
Native Hawaiian or Pacific Islander 49.2/48/26.6/32 54.2/56/25.7/32 54.2/58/26.0/33 57.4/63/24.8/33
mean / median / SD / N, NH SAS grades 4–8 pooled, 2026. Native Hawaiian or Pacific Islander and American Indian or Alaska Native have small N; cells with N < 10 are suppressed. For reference, SGPs uniformly distributed over the integers 1–99 have SD = √((99² − 1) / 12) ≈ 28.6.

Equity gaps at a glance

Horizontal bar chart of mean SGP for each student group, ELA and mathematics pooled, with cohort and baseline bars side by side and one panel per demographic dimension. A dashed line marks 50.

Within-year growth is close to comparable across groups

Group vs. comparison group ELA Mathematics
Gender (Female vs Male) 51.4 vs 48.5 49.9 vs 49.9
English learners 49.1 vs 50.0 49.5 vs 49.9
Students with disabilities 44.1 vs 51.4 45.8 vs 50.9
Economically disadvantaged 46.9 vs 50.9 46.4 vs 51.0
Mean cohort-referenced SGP, NH SAS grades 4–8 pooled, 2026.

Where achievement gaps are large, within-year growth is often not. Cohort SGPs near 50 for both members of a pair mean the gap is being maintained rather than widened this year.

Target attainment

What the target actually is

“Catch up” and “keep up” mean nothing without a stated horizon. In New Hampshire the NH SAS target is proficiency by the end of the grades 3–8 progression, so the horizon shrinks as students approach grade 8.

Grade Horizon to target Catch-up rate Keep-up rate
Grade 4 3 years 36.0% 65.4%
Grade 5 3 years 34.2% 63.7%
Grade 6 2 years 28.8% 68.1%
Grade 7 1 year 23.3% 76.4%
Grade 8 same year 13.8% 80.7%
Cohort targets, NH SAS, ELA and mathematics pooled, 2026. Horizon is the median number of years to the target.

As the horizon shrinks, catch-up rates fall and keep-up rates rise. Most of the apparent grade gradient is the target, not the students.

Catch-up / keep-up attainment (NH SAS, 2026)

Bar chart of the percentage of students on track to catch up or keep up, by grade and horizon to target, one panel per subject. Bars show 95% intervals and the number of students.

Target attainment by prior achievement

16%
of students who were Below Proficient last year
are on track to catch up
N = 27,027
82%
of students who were Above Proficient last year
are on track to keep up
N = 22,775
These are different students judged against different targets, so the two figures are not a gap. NH SAS, ELA and mathematics pooled, 2026.

Students entering below proficient face the steepest catch-up targets and attain them least often — the central equity concern for accountability use of these growth results.

WIDA ACCESS for New Hampshire

What was run: WIDA ACCESS

English learners in New Hampshire take ACCESS for ELLs. The SGP analyses treat the Overall ELP composite as READING and ask whether students are on track to the state’s exit standard.

  • Exit / proficiency criterion: WIDA Level 4.5 (in SGPstateData and in the prep scripts since 2019). Fine-grained scores 4.5–4.9 are collapsed to Level 4.5.
  • Cohort SGPs for 2026 were computed on the new ACCESS 3.0 scale.
  • Baseline SGPs and baseline catch-up / keep-up targets were computed on the old ACCESS 2.0 scale, against the existing 4.5 cuts. Those are the headline on-track results in this section.
  • Targets run 1 to 6 years forward. The rates below use the 6-year horizon — the longest window present in the object.

2026 is a transition year

WIDA_NH does not yet carry the consortium’s 2026 transition metadata (new cuts, knots, linkages). Cohort targets computed on the new scale against old-scale cuts are not on the same footing as the baseline targets. They are shown only as a discussion item.

ACCESS participation by grade, 2026

Grade N tested N with cohort SGP N with baseline SGP
K 506 0 0
1 557 443 443
2 611 511 511
3 602 522 522
4 631 564 564
5 398 337 337
6 352 286 286
7 389 318 318
8 370 320 320
9 420 336 336
10 353 286 286
11 321 256 256
12 302 225 225
WIDA ACCESS Overall ELP, 2026. Kindergarten is shown as K. Cells with N < 10 are suppressed later in the on-track tables, not here.

Percent at or above Level 4.5 over time

Line charts, one per grade from K to 12, of the percentage of tested students at or above WIDA Level 4.5 in each year.

2026 WIDA SGP at a glance

Grade Cohort (new scale) Baseline (old scale)
K — —
1 49.1/49/443 31.6/24/443
2 49.3/49/511 63.8/70/511
3 48.4/48/522 51.8/53/522
4 49.6/50/564 59.2/64/564
5 49.3/50/337 54.8/57/337
6 48.1/47/286 59.7/63/286
7 49.1/50/318 58.5/65/318
8 49.1/48/320 58.3/64/320
9 49.0/51/336 56.9/60/336
10 48.4/48/286 53.7/60/286
11 49.4/50/256 53.9/54/256
12 47.9/49/225 51.4/54/225
mean / median / N, WIDA ACCESS Overall ELP, 2026. Baseline cells are the headline growth results for this transition year. Cells with N < 10 are suppressed.

Baseline SGP trend, WIDA ACCESS

Line charts, one per grade from 1 to 8, of mean baseline SGP on WIDA ACCESS by year. A dashed line at 50 marks the historical ACCESS 2.0 rate of learning.

The 2026 scale transition

Grade N Mean (new − old) Mean |old − linked(new)| 4.5 cut
old → new		 % ≥ 4.5
(matched cut)		 % ≥ 4.5
old cut on new score
K 506 -10.8 0.00 309 → 283 6.5% 2.8%
1 557 +36.3 0.00 330 → 374 2.3% 13.6%
2 611 +25.0 0.00 344 → 378 9.3% 42.2%
3 602 +28.8 0.00 356 → 394 13.0% 51.5%
4 631 +13.5 0.00 366 → 381 37.7% 56.9%
5 398 +12.7 0.00 374 → 391 24.4% 43.7%
6 352 +34.3 0.00 382 → 430 7.1% 46.0%
7 389 +37.5 0.00 389 → 439 14.1% 52.7%
8 370 +39.3 0.00 395 → 448 13.0% 55.7%
9 420 +40.1 0.00 402 → 457 12.9% 55.0%
10 353 +39.7 0.00 408 → 465 6.8% 53.3%
11 321 +40.6 0.00 413 → 471 9.0% 49.8%
12 302 +39.0 0.13 418 → 477 5.0% 43.0%
Both composites are scored 100–600; that is a shared range, not a shared scale. The vendor old-scale composite is the consortium NEW_to_OLD concordance of the new composite (largest grade mean gap 0.13 points), not a separate measurement. The new-scale 4.5 cut is the smallest new score that the concordance links to the old cut. The matched-cut rate is identical on either scale and agrees with the vendor-reported level for 5,812 of 5,812 tested students. The last column applies old cuts to new scores and is not a valid result. The concordance is the one on generic WIDA in SGPstateData; WIDA_NH does not yet carry it.

Levels carry across the transition; raw composites do not. WIDA’s interim 2026 cuts are the concordance images of the old cuts, so a 2026 Level 4.5 means the same thing as a 2025 Level 4.5 to the extent the concordance holds. Mean (new − old) is not growth, and the last column is the mistake to avoid.

One concordance per ACCESS grade cluster

Line chart of the 2026 consortium concordance: for each grade cluster, the linked ACCESS 2.0 composite plotted against the new ACCESS 3.0 composite, with a dashed identity line. Flat runs mark floors and ceilings.

The 4.5 exit cut on each scale

Line chart of the Level 4.5 exit cut by grade, K to 12, on the old scale and its image on the new scale. The old cuts rise every grade; the new-scale cuts do not, and circled points mark a new cut lower than the grade before. Shaded bands mark the concordance grade clusters.
The grade 3→4 dip is not new. On the old scale, grade 4 had the highest share at or above 4.5 in 9 of 9 years from 2017 to 2025 (25–42%, against 9–20% in grade 3). On the new scale, that long-standing seam between the grade 2–3 and 4–5 concordance tables shows up as a cut that falls.

What-if: standard setting smooths the new cuts

Line chart of the new-scale Level 4.5 cut by grade, K to 12, under two smoothing methods.

Grade N Interim Isotonic Concave
1 557 2.3% 2.3% 4.1%
2 611 9.3% 9.3% 10.1%
3 602 13.0% 17.9% 17.9%
4 631 37.7% 31.7% 21.7%
5 398 24.4% 24.4% 10.6%
6 352 7.1% 7.1% 10.2%
7 389 14.1% 14.1% 16.5%
8 370 13.0% 13.0% 14.6%
9 420 12.9% 12.9% 13.8%
% of 2026 students at or above 4.5 on the new composite under each set of cuts. Red cells move. Unchanged under both smoothers: grades K, 10, 11, 12. A what-if, not a prediction of WIDA’s cuts.

Baseline growth and on-track results do not move: they use the old scale and the old 4.5 cuts. What moves is a 2026 exit decision made on the new composite, mostly in grades 3–5.

On track to Level 4.5 by grade (baseline, 6-year)

Bar chart, by grade, of the percentage of students below WIDA Level 4.5 who are on track to reach it within six years under baseline norms, with 95% intervals.

On track by current WIDA level

Current levelCatch-up rateN
L10.4%471
L28.7%835
L338.5%1,719
L469.4%756
L4.5100.0%383
L5100.0%212
L6100.0%28
Baseline 6-year catch-up, old scale, 2026. The 6-year CUKU column codes students who started the window below 4.5; students already at the exit standard do not receive a 6-year Keep Up status. Cells with N < 10 are suppressed.

Years to the Level 4.5 target

Bar charts, one per grade from 1 to 8, of the distribution of years needed to reach WIDA Level 4.5 under the baseline six-year configuration.

Cohort vs. baseline on-track — a discussion item

Do not treat these two columns as a trend
Baseline catch-up projects on the old scale and compares with the current 4.5 cuts. Cohort catch-up feeds each 2026 new-scale score into 2026 matrices whose prior-score slots were fit on old-scale scores, then compares the projections with the same old-scale cuts. The difference is a scale-transition artefact, not a finding.
Grade Baseline catch-up (old scale) Cohort catch-up (new scale)
1 28.0% (N = 443) 84.2% (N = 443)
2 73.2% (N = 511) 87.1% (N = 511)
3 57.7% (N = 522) 78.7% (N = 522)
4 62.4% (N = 564) 77.1% (N = 564)
5 40.7% (N = 337) 77.4% (N = 337)
6 43.7% (N = 286) 77.6% (N = 286)
7 46.2% (N = 318) 81.8% (N = 318)
8 41.2% (N = 320) 85.3% (N = 320)
9 33.6% (N = 336) 83.9% (N = 336)
10 13.3% (N = 286) 70.3% (N = 286)
11 12.5% (N = 256) 62.1% (N = 256)
12 4.4% (N = 225) 4.4% (N = 225)
Catch Up only, 6-year horizon, 2026. Keep-up rates omitted here so the contrast is readable.

WIDA ACCESS — baseline model fit, 2026

Grade 3
Grade 3
Grade 4
Grade 4
Grade 5
Grade 5
Grade 6
Grade 6
Grade 7
Grade 7
Grade 8
Grade 8

Summary

  • NH SAS cohort SGPs show even conditional coverage across grades 4–8 in ELA and mathematics — suitable for accountability calculations.
  • NH SAS baseline SGPs are the diagnostic lens on recovery relative to 2018–19. Grades flagged more than two points from 50 are listed on the baseline-results slide.
  • Within-year growth is close to comparable across student groups; where it is not, that pair is the priority.
  • Catch-up / keep-up rates are not comparable across grades without their target horizon, which shrinks toward grade 8.
  • WIDA ACCESS exit remains WIDA Level 4.5. Headline on-track results are baseline-referenced and computed on the old composite scale.
  • 2026 is an ACCESS 3.0 transition year. Cohort targets on the new scale against old cuts should not be reported as a trend. Whether WIDA_NH adopts the consortium’s 2026 transition metadata is a TAC question.

Questions for the TAC

Where the committee’s guidance is needed

  1. How long should New Hampshire carry the old-scale WIDA baseline after ACCESS 3.0? The 2026 split run is possible only because both composites are on the file.
  2. Should WIDA_NH adopt the consortium 2026 transition metadata now (cuts, knots, linkages already on the generic WIDA entry), or wait until new standards are set? If those standards smooth the grade-cluster seams, grade 3–5 exit rates will move.
  3. Should 2026 cohort on-track results be withheld from public reporting, the way a contradicted baseline series would be, until the new-scale cuts exist?
  4. Should the 2018–19 NH SAS baseline continue as the monitoring series, or is it time to discuss a dual track — a fixed recovery series plus a re-established reporting norm?

Thank you

New Hampshire Technical Advisory Committee

New Hampshire 2026 SGP Results

Damian W. Betebenner · National Center for the Improvement of Educational Assessment