Every published figure on this site is the Arkansas Division of Elementary and Secondary Education's own published figure at that grain. District, county, co-operative and state figures are read from ADE's own district, County, COOP and State sheets — they are not totals we produced by adding up schools. That matters: about half of all school-level cells are suppressed for small group size, so a total built by adding up the visible schools would be a partial sum presented as a complete one.
What is computed here, and labelled as computed: year-over-year changes, and the median of a school's peer group. Both are arithmetic over published figures.
| ATLAS year | Shown as | ADE's own stage name | School year |
|---|---|---|---|
| 2024 | Final | ATLAS Summary Post-Appeals Scores Spring 2024 | 2023-24 |
| 2025 | Final | ATLAS Summary Post-Corrections Scores Spring 2025 (XLSX) | 2024-25 |
| 2026 | Preliminary | ATLAS Summary Pre-Appeals Scores Spring 2026 (XLSX) | 2025-26 |
ADE renamed its final release stage between years: Spring 2024's is "Post-Appeals" and Spring 2025's is "Post-Corrections". They mean the same thing — the results after appeals and corrections have run — so both are shown as Final. Displaying both names side by side would invite readers to think the two years differ in some way they do not. ADE's exact stage name is kept in the table above and in every page's downloadable data.
ATLAS is administered in the spring of each school year. Year labels on this site are that spring — "2026" is the spring 2026 administration, which falls in the 2025-26 school year — so the word "spring" is left off the charts and tables rather than repeated on every label.
ADE hides results for small groups of students. We show that as suppression and never estimate the hidden value, never substitute zero, and never leave a blank that could be read as zero.
| What you see | What it means |
|---|---|
| suppressed (N<10) | Fewer than 10 students; ADE publishes nothing for the cell. |
| suppressed (RV) | ADE's own suppression code: the number tested is real, the results are withheld. |
| not published this year | Spring 2024 and 2025 do not publish per-level student counts at all. Absent is not the same as hidden. |
| counts suppressed | New in Spring 2026: ADE publishes the percentages and the number tested but hides the per-level counts. |
| — | The subject is not tested at that grade anywhere in Arkansas. |
In Spring 2026 the hidden per-level counts could be reconstructed arithmetically from the published percentages. We deliberately do not do this. A cell ADE chose to suppress stays suppressed here.
Where Spring 2026 publishes per-level student counts, every percentage on this site is derived from those counts, so the level bars always sum to 100 and always agree with the headline figure. ADE's own published percentages disagree with its own published counts in 396 of 6,620 cells — always by dividing by a slightly larger denominator than the published number tested, by up to 1.1 percentage points. Where counts are not published, ADE's percentages are used exactly as published. Each figure records which of the two it is.
Spring 2026 figures come from ADE's Pre-Appeals release and can still change. At that stage ADE's statewide total runs above the sum of its own districts by between +0.006% and +0.184% depending on subject (ELA: 415 students out of 283,156), because some test records are not yet attributed to a school. Spring 2024 and Spring 2025, both final releases, reconcile exactly with zero gap. Where this site shows a state figure it shows ADE's own State sheet figure, never a sum of districts, so one sourcing rule holds everywhere.
A percentage over fewer than 30 students moves several points on one student's result. Those figures are shown with the number tested attached and a caution badge, are left out of statewide distributions and peer medians, and are not placed in a quartile band. Charts state how many entities were excluded rather than dropping them silently.
The headline trend is cross-sectional: it is the same grade each year, made up of different students. A school's grade-4 result rising does not by itself mean any child improved. The cohort view follows one group forward instead — grade 4 to grade 5 to grade 6 — which is the closer thing to progress, though students still move in and out of a school between years.
Twelve nearest schools inside the same grade-span band, by rank-normalized distance over four equally weighted dimensions: economically-disadvantaged percentage, log enrollment, English-learner percentage and special-education percentage. Two schools with the same poverty rate and the same size can serve very different populations — one with a third of its students learning English, one with almost none — and a peer set that ignores that is comparing the wrong things. District pages use the same formula, with the grade-span band replaced by whether the district runs kindergarten through grade 12.
The band is a filter, not a fifth distance term: as a soft term it would pair a K-5 with a 9-12 whenever the other dimensions happened to line up. An entity missing any one of the four dimensions gets no peer panel and a note saying so, rather than a peer set built on whichever dimensions happen to be present.
Peer medians are computed from ADE's published figures and labelled as computed. Peers below 30 tested students are excluded and counted.
Arkansas data carries two, and they are not interchangeable. The site displays the audited Child Nutrition free-and-reduced figure and matches peers on MSI's ESSA economically-disadvantaged percentage, which covers more schools (1,003 against 885), so one consistent definition runs through the whole comparison set. Both are labelled wherever they appear.
At school grain the two turn out to agree exactly on 924 of 929 schools — MSI's school-level economically-disadvantaged figure appears simply to be the Child Nutrition free-and-reduced figure. At district grain they genuinely differ (Green Forest: 85.4% against 74.9%). So the split between what is shown and what is matched on rests on coverage, not on the two being different definitions at school level.
Districts operating community eligibility (CEP) report all students as eligible, so their free-and-reduced figure is flagged and is not comparable with a non-CEP district's.
Both come from My School Info, at school and district grain, for the school year matching the latest ATLAS administration. Arkansas also publishes a special-education count through its special-education office, taken as a December 1 child count; that figure and My School Info's percentage do not agree, by an average of about one percentage point and occasionally much more, because the two use different timing and different denominators. The site uses My School Info's figure throughout so that one definition runs across every school and district, and says so here rather than presenting a blend of the two.
Every figure on this site comes from ADE. The county and co-operative outlines do not, because ADE publishes no boundary file: they are U.S. Census Bureau cartographic boundary files, vintage 2025, used for their geometry only. No federal figure of any kind is used as data anywhere on this site — the shapes come from the Census, the numbers colouring them are ADE's own.
Education service co-operatives are groupings of school districts, not of counties, so a co-operative's area is its member districts' boundaries combined, using ADE's own Master List of which district belongs to which co-operative. 234 of 264 districts have a published boundary. The remaining 30 are open-enrollment charters and state-operated schools, which have no geographic service area for the Census to publish; they are counted in their co-operative's ADE figure but add no area to the map.
Boundaries carry no LEA number — Arkansas's own district numbering appears in no federal file — so each polygon is matched to a district by name and then confirmed against the county ADE assigns that district. That check is what tells Arkansas's two Harmony Groves, two Lakesides and three West/Westsides apart. Anything that could not be resolved to exactly one district is reported as unmatched rather than assigned to a best guess.
A county with no computable change is drawn hatched, never in a pale colour that could be read as "roughly no change", and a map never carries a figure on its own: the same counties and co-operatives are listed underneath in text, with their numbers, which is also what prints and what a reader who does not distinguish these colours uses.
Every district, mapped is its own page rather than a section of the statewide one because it carries every subject at every grade for all three years — about 800 KB of published figures, which the statewide page would download to draw a map it does not have. Every value on it is ADE's own published district figure at the selected grade, and the change views are the difference between two of them, shown only where both years published a figure over 30 tested students at that grade.
The three single-year views share one fixed colour scale — breaks at 10, 20, 30, 40 and 50 percent — rather than a scale fitted to each year. A per-year scale would repaint the whole state every time the year changed and show nothing about whether anything actually moved. For the same reason the change views use fixed percentage-point breaks rather than quantiles of the districts in view.
The ranked table carries all 264 districts, including the 30 with no boundary, and any district that published nothing for the selected subject and grade — listed with the reason, unranked, rather than left out. The rank number is the district's position on whichever column is sorted, counting from the highest value down, and it does not change when the sort direction is reversed: reversing a sort re-orders the rows, it does not renumber the districts. It is a position within one selection, not a standing.
It is not a ranking. There is no composite score and no letter grade, and no precise statewide rank is reported — only a quartile band. ATLAS results are never spliced together with ACT Aspire or ACT results: they are different tests on different scales, and a line drawn across them would be meaningless.
This is an independent site. It is not affiliated with, endorsed by, or operated by the Arkansas Division of Elementary and Secondary Education. It is built entirely from data ADE publishes.
© 2026 Joel Rush / Sidetown Labs. All rights reserved.
Some uses need no permission at all. Screenshots, excerpts and printed
copies of any page, chart or table are welcome — with attribution to
arkansas-atlas.sidetownlabs.com — for schools, school districts, education service
cooperatives, school boards, press, and academic or policy research. Linking to any page is
always fine. This site exists to be used in a board packet or a news story, and that use is
pre-authorized so nobody has to ask first.
Anything else — republishing substantial portions, redistributing the site's files, offering a derived or modified version, or any commercial use — needs written permission: joelrush@gmail.com. Permission is usually a short email away; the point of asking is that a copy of this site elsewhere goes stale, and a stale copy of assessment figures is worse than none.
The underlying figures are public records published by the Arkansas Division of Elementary and Secondary Education, and no claim is made over the facts themselves. What is covered here is this site's text, code, design, charts and compilation — the arrangement, not the numbers. ADE's own published workbooks remain available from ADE, and every figure on this site is downloadable from the page that shows it.
Citation: Rush, Joel. Arkansas ATLAS — three-year trends. Sidetown Labs, 2026. https://arkansas-atlas.sidetownlabs.com/.