Considerations for Illinois SGP Analyses

During the Transition from SAT/PSAT to ACT/PreACT

Damian W. Betebenner

National Center for the Improvement of Educational Assessment

Illinois TAC Meeting · May 21, 2026

Illinois SGP Analyses

Considerations During the Transition to ACT

Cohort- and Baseline-Referenced SGPs Across the SAT/PSAT → ACT/PreACT Vendor Change

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

Illinois TAC · May 21, 2026

Roadmap

Background

  • The current Illinois growth workflow
  • The 2025 high-school assessment transition
  • Two simultaneous policy changes to keep separate

The SGP Distinction

  • Cohort-referenced vs baseline-referenced SGPs
  • Why baseline matrices do not travel across vendors

Cohort-Referenced Transition Map

  • 2025 · 2026 · 2027 grade-9/10/11 patterns
  • Recurring vs transitional sequences

Baseline-Referenced Implications

  • Calibration timing by grade span
  • Three baseline design choices

Performance-Level Reset

  • What is and is not affected

Implementation 2026–2027

  • Configuration, knots, SIMEX, linking

Recommendations & Decisions

  • Reporting approach, ISBE decision points
  • Bottom line and timing roadmap

Purpose

This memorandum lays out technical and reporting considerations for Illinois SGP analyses in ELA and mathematics for grades 4 through 11 during the transition from the SAT/PSAT high-school system to the ACT/PreACT high-school system.

The operational picture

  • Grades 4–8 IAR analyses can continue under the established cohort + baseline workflow
  • High school enters a transition period beginning spring 2025
  • Grade 11: ACT with Writing
  • Grade 10: PreACT Secure
  • Grade 9: PreACT 9 Secure

The technical question

The discussion distinguishes throughout between cohort-referenced and baseline-referenced SGPs because the two approaches are affected very differently by the vendor transition.

Cohort-referenced SGPs survive the transition with caveats. Baseline-referenced high-school SGPs probably require new ACT-era coefficient matrices before they can be reported operationally (equating NEW scale to OLD can possibly be used as a workaround).

Background

The Transition

What changed, when, and why both vendor and policy changes are in play.

The Current Illinois Growth Workflow

Grades 4–8 IAR — established and stable.

  • Current production script (PARCC_IL_SGP_2024_2025.2.R) loads three panel years (2022–2023, 2023–2024, 2024–2025)
  • Configures ELA and mathematics for grade sequences:
3-4
3-4-5
4-5-6
5-6-7
6-7-8
  • Runs abcSGP requesting:
    • sgp.percentiles (cohort-referenced)
    • sgp.percentiles.baseline (baseline-referenced)
    • sgp.projections and sgp.projections.lagged
    • sgp.projections.baseline and sgp.projections.lagged.baseline
    • SIMEX measurement-error correction

What this means going into the transition

  • Grades 4–8 stay on rails: same scale family, same baseline matrices, same projection workflow.
  • The interesting questions live entirely in grades 9–11.

The grade-band split is the cleanest organizing principle for the rest of the discussion. Grades 4–8 are an IAR-to-IAR problem; grades 9–11 are a vendor-transition problem.

The Spring 2025 High-School Vendor Change

Before (through 2024)

Grade Operational assessment
11 SAT with Essay
10 PSAT 10
9 PSAT 8/9

Beginning spring 2025

Grade Operational assessment
11 ACT with Writing
10 PreACT Secure
9 PreACT 9 Secure

The 2026 Pre-ID documentation continues the ACT/PreACT structure for grades 9, 10, and 11.

Why this is not just a cosmetic change

  • The scale changes by grade — see next slide
  • The conditional reference distribution changes year over year as ACT-family priors gradually replace College Board priors
  • The first year in which a grade-11 student has an ACT-family current score and both priors in the ACT family is 2027

2027 is the calibration anchor for any future ACT-era grade-11 baseline matrices.

ACT-Family Scales by Grade

Grade Assessment Scale
9 PreACT 9 Secure 1–32
10 PreACT Secure 1–35
11 ACT with Writing 1–36

Implications for the SGP workflow

  • SGP estimation does not require a vertical scale across grades — conditional percentiles are well-defined within grade
  • It does require correct, grade-specific knots and boundaries on each current-score scale
  • ELA and mathematics score distributions need direct validation before production

The grade-specific scales do not prevent SGP calculation. They do require careful, content- and grade-specific scale validation in SGPstateData[["IL"]] before any production run.

The 2025 production script already shows the right pattern in its science knots/boundaries block — that pattern needs to be extended to ELA and mathematics on the ACT-family scales.

The SGP Distinction

What Kind of SGP?

Cohort-referenced and baseline-referenced SGPs answer different questions.

What Is an SGP?

Student Growth Percentile

Let Yg,tY_{g,t} denote a student’s current-year scale score in grade gg and year tt, and let 𝐇g,t\mathbf{H}_{g,t} denote that student’s available prior achievement history. An SGP is a percentile rank of the current score within an estimated conditional distribution:

SGPg,t=F̂Yg,t∣𝐇g,t(Yg,t∣𝐇g,t)\mathrm{SGP}_{g,t} = \widehat{F}_{Y_{g,t}\mid \mathbf{H}_{g,t}}(Y_{g,t}\mid \mathbf{H}_{g,t})

The critical distinction is the reference distribution used for F̂\widehat{F}.

  • Cohort-referenced: F̂\widehat{F} estimated from the current operational cohort
  • Baseline-referenced: F̂\widehat{F} fixed by saved coefficient matrices from a historical reference group

Same percentile machinery, different reference distribution — and that single difference is what makes the vendor transition affect the two SGP types so differently.

Cohort-Referenced SGPs

The question answered

Among students in this year’s statewide cohort with similar prior achievement histories, where does this student’s current achievement fall?

Why this survives the vendor transition

  • The current cohort itself supplies the conditional reference distribution
  • No vertical scale is required across grades or vendors
  • Prior IAR, PSAT, or SAT scores can be used as conditioning variables alongside current ACT/PreACT scores
  • 2025 is the first ACT-family year; the cohort SGP describes a student’s standing relative to other 2025 grade-11 students with similar PSAT 10 and PSAT 8/9 histories

For transition years, cohort-referenced SGPs are the most defensible operational growth metric.

The catch

The transition is part of the estimand. A 2025 grade-11 cohort SGP is not directly comparable to a future 2028 grade-11 cohort SGP whose reference cohort has fully ACT-family priors. Public-facing language should consider labeling transition-year cohort SGPs as such.

Baseline-Referenced SGPs

The question answered

Where does this student’s current achievement fall relative to the fixed baseline distribution for students with similar prior achievement histories?

How it is implemented

A baseline SGP is anchored to saved coefficient matrices for a specific combination of:

  • state · content area · grade sequence
  • score scale · knots and boundaries
  • measurement-error correction design

Baseline SGPs are valuable only when the saved matrices correspond to the same operational assessment design being used for current students.

Why IAR baseline matrices do not transfer to ACT/PreACT

  • Different score scale (ACT 1–36 vs IAR scale)
  • Different prior-score family (College Board / IAR vs ACT family)
  • Different score precision and CSEM definitions
  • Different conditional distribution shape

Category Error

Applying IAR baseline matrices to ACT/PreACT current scores would be a category error. Applying transition-year matrices to later stable ACT-era sequences would also be problematic if the current/prior assessment pattern differs.

Cohort-Referenced SGPs

The Transition Map

Year-by-year, grade-by-grade — what’s current, what’s prior, and what it means.

2025 — First ACT-Family Year

Grade Current First prior Second prior
9 PreACT 9 Secure 2024 IAR grade 8 2023 IAR grade 7
10 PreACT Secure 2024 PSAT 8/9 grade 9 2023 IAR grade 8
11 ACT with Writing 2024 PSAT 10 grade 10 2023 PSAT 8/9 grade 9

Reading the row

  • Grade 9 is the recurring IAR-to-PreACT-9-Secure pattern that we will see every year going forward
  • Grade 10 and grade 11 are transition sequences with College Board priors
  • All three current scores are ACT-family for the first time

2025 cohort-referenced HS SGPs are technically feasible — but they describe a one-of-a-kind cohort with mixed-vendor priors. They should be reported with explicit transition-year labeling.

2026 — ACT-Family Current and First Prior

Grade Current First prior Second prior
9 PreACT 9 Secure 2025 IAR grade 8 2024 IAR grade 7
10 PreACT Secure 2025 PreACT 9 Secure gr 9 2024 IAR grade 8
11 ACT with Writing 2025 PreACT Secure gr 10 2024 PSAT 8/9 grade 9

Reading the row

  • Grade 10 is now ACT-family in current and first prior; second prior is grade 8 IAR — this is the recurring grade-10 sequence
  • Grade 11 is ACT-family in current and first prior, but the second prior is still PSAT 8/9 — a one-prior-residual transition year
  • A one-prior 2026 grade-11 model would be ACT-family throughout — but would not be directly comparable to a two-prior model

2026 is the year the grade-11 model decision matters most: two priors keeps the College Board residual; one prior is fully ACT-family but a different model.

2027 — First Full ACT-Family Grade-11 Sequence

Grade Current First prior Second prior
9 PreACT 9 Secure 2026 IAR grade 8 2025 IAR grade 7
10 PreACT Secure 2026 PreACT 9 Secure gr 9 2025 IAR grade 8
11 ACT with Writing 2026 PreACT Secure gr 10 2025 PreACT 9 Secure gr 9

Reading the row

  • All three current scores and all priors are in the ACT family for the first time
  • This is the first natural calibration year for a two-prior ACT-era grade-11 baseline
  • Grades 9 and 10 are on their recurring IAR-to-ACT-family sequences

2027 is the calibration anchor. Operational use of a full grade-11 ACT-era baseline should not occur before 2028, after technical review of the calibration year.

Implications by Grade Band

Grades 4–8 (IAR)

  • 2026 and later remain IAR-to-IAR analyses
  • Continue the standard grade sequences 3-4, 3-4-5, 4-5-6, 5-6-7, 6-7-8
  • Cohort + baseline workflows both stable

Grade 9

  • IAR-to-PreACT-9-Secure is a recurring grade-span transition, not a temporary vendor transition
  • Grade 9 SGPs will always condition on middle-school IAR history under the current rule
  • A separate prior-data rule would change this

Grade 10

  • 2026 is the first year with ACT-family current and first-prior scores
  • Second prior is grade 8 IAR — part of the recurring grade-10 sequence
  • 2026 is therefore the natural first candidate calibration year for a recurring grade-10 ACT-era baseline

Grade 11

  • 2026: ACT-family current and first prior; PSAT 8/9 still in the second-prior slot under a two-prior model
  • 2027: ACT with Writing, PreACT Secure, PreACT 9 Secure all present in the same longitudinal sequence

Recurring vs Transitional Sequences

A useful organizing distinction across the three transition years:

Recurring sequences — will repeat every year going forward

  • Grades 4–8 IAR-to-IAR (no change)
  • Grade 9: IAR grades 7, 8 → PreACT 9 Secure
  • Grade 10 (from 2026): IAR grade 8 → PreACT 9 Secure → PreACT Secure
  • Grade 11 (from 2027): PreACT 9 Secure → PreACT Secure → ACT with Writing

Transitional sequences — appear only during the vendor changeover

  • Grade 10 in 2025: PSAT 8/9 → PreACT Secure
  • Grade 11 in 2025–2026: PSAT 8/9 / PSAT 10 → ACT with Writing

Calibration follows the recurring sequences, not the transitional ones. Transitional years can be reported cohort-referenced; baseline calibration should anchor to the first stable year for each recurring sequence.

This is why grade-9 and grade-10 baseline calibration could become possible earlier than grade-11 baseline calibration — they reach a stable recurring sequence sooner.

Baseline-Referenced SGPs

A Separate Design Problem

Calibration timing, design choices, and the operational default during transition.

Calibration Timing by Grade Span

Grade span Recurring assessment sequence First stable calibration year Baseline recommendation
Grades 4–8 IAR current with IAR priors Already available Continue cohort + baseline; update projections to new benchmarks
Grade 9 PreACT 9 Secure with IAR gr 8 (and gr 7) priors 2025, but also first ACT-family year Cohort during transition; ACT-era baseline only after 2025–2027 distributions show stability
Grade 10 PreACT Secure with PreACT 9 Secure (gr 9) and IAR gr 8 priors 2026 Calibrate from 2026 (or pooled 2026–2027); operational use no earlier than 2027
Grade 11 ACT with Writing with PreACT Secure (gr 10) and PreACT 9 Secure (gr 9) priors 2027 Full two-prior baseline not operational before 2028; one-prior 2026 baseline only as a distinct, explicitly labeled model

Different grades become stable in different years. A pooled or grade-specific calibration design is therefore unavoidable.

Three Baseline Design Choices

Single baseline year

  • Easiest to explain
  • Closest to the current fixed-baseline interpretation
  • Risk: sensitive to first-year administration effects if the chosen year is too early
  • Defensible only after technical evaluation of distributional stability

Pooled baseline period

  • Pool 2026–2027, or 2025–2027 where the assessment sequence is appropriate
  • Stabilizes coefficient estimation
  • Pooling design must be grade-specific because grades reach stability in different years
  • Stronger fixed-reference interpretation than rolling

Rolling reference

  • Improves recency
  • Weakens the fixed-reference interpretation
  • Better understood as a smoothed cohort reference than a traditional baseline
  • Reasonable as a transitional reporting choice — less clearly a “baseline”

The choice is not just statistical — it is interpretive. Each option implies a different answer to “what does a 50 mean?”

Default Recommendation Until ACT-Era Matrices Exist

For accountability reporting in 2026 and 2027:

  1. Grades 4–8: Continue cohort + baseline-referenced ELA and mathematics SGPs using the established IAR workflow and existing Illinois baseline matrices.

  2. Grades 9, 10, 11: Report only cohort-referenced SGPs.

  3. Do not report baseline-referenced HS SGPs until ACT-era matrices have been:

    • created for the recurring grade sequences
    • reviewed for distributional stability
    • approved for the intended accountability use

Defensible default: keep the grades 4–8 baseline workflow, add cohort-only HS analyses with explicit transition labels, and defer HS baseline reporting until ACT-era matrices are deliberately calibrated.

This is the same posture the 2025 production script already supports — it requests baseline outputs only for the IAR grade sequences. Extending to HS only requires adding cohort-referenced grade-9, grade-10, and grade-11 sequences without enabling baseline flags.

SIMEX, Linking, and Participation

SIMEX measurement-error correction

  • The 2025 script already requests both uncorrected and SIMEX-corrected SGPs
  • For grades 4–8: continue if IAR score and CSEM fields remain consistent
  • For HS: SIMEX requires defensible CSEM values for ACT/PreACT score variables
  • If ACT/PreACT CSEM values are unavailable or not comparable to IAR CSEM, run uncorrected cohort SGPs for HS or complete a separate technical evaluation first

Student linking and participation rules

  • Vendor systems, Pre-ID processes, and assessment records all changed during the transition
  • The 2025 script already contains an Illinois-specific data-cleaning step that renames student identifier fields in prior-year files — audit this across ACT/PreACT, PSAT/SAT, and IAR files
  • 2026 Pre-ID: SIS enrollment auto-includes students except where IDEA + DLM-AA = Yes
  • For growth, the question is operational accountability score validity, not college reportability

Three independent risk surfaces — SIMEX, linking, and participation rules — each warrant a discrete go/no-go check before HS production.

Recommendations

What to Report, How to Label It

An operational posture for 2026 reporting and the decisions Illinois needs to make.

Suggested Public-Interpretation Language

High school SGPs are cohort-referenced measures calculated from the current year’s statewide assessment results and available prior state assessment scores.

Because Illinois transitioned from SAT/PSAT to ACT/PreACT beginning in spring 2025, high school SGPs for 2025 and some 2026 grade sequences reflect transition-year assessment histories.

These results describe student growth relative to academic peers in the same statewide cohort; they are not baseline-referenced SGPs unless explicitly identified as such.

Suggested verbatim language for ISBE public reporting during 2025 and 2026 high-school growth releases.

Bottom Line — Timing Roadmap

Year-by-year operational picture

Year Grades 4–8 Grades 9–11
2025 Cohort + baseline (IAR) Cohort only — first ACT-family year
2026 Cohort + baseline (IAR) Cohort only; 2026 candidate calibration for grade 10
2027 Cohort + baseline (IAR) Cohort only; first full ACT-family grade-11 sequence — calibration anchor
2028 Cohort + baseline (IAR) Earliest prudent operational use of a two-prior grade-11 ACT-era baseline

The ACT transition does not prevent Illinois from calculating high-school cohort-referenced SGPs. It does prevent a simple extension of the existing baseline-referenced IAR workflow to high school.

The strongest technical course

  • Continue grades 4–8 cohort + baseline analyses (with updated baseline matrices)
  • Add cohort-referenced HS analyses with explicit transition labels
  • Delay HS baseline-referenced reporting until ACT-era matrices are calibrated, validated, and approved
  • Under a two-prior model, 2028 is the earliest prudent year for operational HS baseline use

Considerations for Illinois SGP Analyses

During the Transition from SAT/PSAT to ACT/PreACT

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

Thank you · Discussion

Illinois TAC · May 21, 2026