DepEd’s public open datasets are 17 files with 2,216 columns, covering enrollment from SY 2017-18 to SY 2025-26, SY 2023-24 school inputs, and three assessments. Below are what the files contain, what must be fixed before any number can be trusted, and what the files cannot say however carefully they are read.

The checks are the ones a statistician runs before analysis: duplicates, missing values, impossible values, inconsistent labels, matching across files, and coverage.

The companion report, Philippine basic education in numbers, holds the full data inventory, a dictionary of every column, and the quality report with a count for each issue.

Key findings

  • 17 archives, 17 CSV files, 2,216 columns, one row per school in every file, and zero duplicate school IDs. The six SY 2023-24 school files hold the same 60,167 schools, and the National Achievement Test (NAT) and Early Language, Literacy, and Numeracy Assessment (ELLNA) files match them completely.
  • The personnel, facilities, information and communication technology (ICT), and water, sanitation, and hygiene (WASH) files hold rows for every school but data for public schools only. 47,818 rows have data, though one of them is blank in the personnel and ICT modules, so those modules count 47,817; the 12,349 rows for private, state-university, and overseas schools are blank by design.
  • The three assessment files cover only a “selected set of schools”. They cover five regions, 44 of their 63 divisions, and about half of the public schools inside those divisions, and DepEd has not published the selection rule.
  • The SY 2025-26 release changed its conventions. It mixes text encodings (1,892 lines in Windows-1252 inside a UTF-8 file), uses new labels for sector, management, and annex status, and leaves blanks where earlier years wrote zeros.
  • Impossible values are rare but real. 139 public schools report learners and no teachers, and 267 report no classroom. 316 report more computers than learners, and one school reports 15,699 tablets.
  • Region boundaries moved four times in nine years. ARMM became BARMM, Cotabato City moved to BARMM, the Negros Island Region was created, and Sulu moved to Region IX. Two division names also exist in two regions each, so regional series need constant boundaries keyed on region and division together.

What was audited

The audited files are DepEd’s proactive-release datasets, and they are of three kinds. Nine enrollment files cover SY 2017-18 to SY 2025-26 and come from the Learner Information System (LIS). Five school-characteristics files for SY 2023-24 cover learners, personnel, facilities, ICT, and WASH and utilities. They come from the Basic Education Information System (BEIS) and the National School Building Inventory (NSBI). Three assessment files come from the Bureau of Education Assessment. They are the NAT Grade 6 and ELLNA files for SY 2023-24 and the Comprehensive Rapid Literacy Assessment (CRLA) file for SY 2025-26.

Each archive holds one CSV file and a README. Two technical notes from DepEd’s Policy and Planning Service and its Bureau of Education Assessment define the indicators, the proficiency bands, and the coverage limits.

Duplicate school IDs and keys

Every file has one row per school, keyed on a six-digit school ID, and no file has a duplicate key. In each year, 130 to 250 rows share a school name, municipality, and barangay with another row under a different ID. These are annex schools and same-named public and private pairs, not true duplicates, so none is removed.

Across the nine enrollment files, 66,967 distinct IDs appear, and 55,029 schools are present in every year. A likely ID change is an exiting school whose name and place reappear under a new ID the next year. There are 146 such cases over the eight year-to-year transitions, which agrees with the README’s note that ID changes are “a minority of cases”. Each is counted as an exit plus an entry; DepEd says a mapping is planned.

Missing values and blank cells

Blanks are of three kinds, and the difference matters: a blank treated as zero distorts an average.

Structural blanks. The four public-only modules are blank for every non-public row, as the technical notes state. Level-specific columns are blank when a school does not offer the level; a school with no senior high school has no senior-high classroom count. Columns that did not exist in a year are blank for that year. Examples are the “unique track” senior-high columns before SY 2024-25, the school district before SY 2018-19, annex status before SY 2019-20, and junior-high non-graded counts in SY 2018-19.

Unexpected blanks. Among public schools that offer the level, 669 have no elementary classroom count. 17 report no nationally funded teachers at the elementary level, 595 at the junior high level, and 213 at the senior high level. One to three schools have blank utility fields. These blanks are treated as missing: the school drops out of the affected indicator and is never scored zero.

Blanks in SY 2025-26. That file writes blanks where earlier years wrote zeros. 1,403 schools have every kindergarten-to-Grade 10 cell blank; 1,347 of them are “Purely SHS” schools, for which the blank means zero. 46 schools have every grade cell blank and reported nothing. They are kept as present with missing enrollment; their SY 2024-25 total was about 465 learners. 128 schools report some levels but leave an offered level entirely blank. 89 of those levels were zero the year before, so the blank is read as zero.

Impossible and implausible values

No file has a negative count, a mean percentage score outside 0 to 100, or a share above 100%. The CRLA level counts sum exactly to the test-taker count in every school. The CRLA README calls its share columns percentages, but they are proportions from 0 to 1.

The implausible values are the ones a planner should know about. 139 public schools report learners and no teachers; their ratio is undefined and is excluded. 7 report more teachers than learners, and 267 report learners and no instructional classroom. 1,362 report no functional toilet bowl (1,360 of them have learners), which is kept as a real shortage.

Several implausible values are in the ICT counts. 316 schools report more academic computers than learners, 108 ICT cells exceed 1,000, and one school reports 15,699 tablets. For ratios, computer counts are capped at the 99th percentile. 192 schools report more than 10,000 learners in some year; these are kept, because large urban schools exist.

Test-taker counts can exceed enrollment. ELLNA test-takers exceed Grade 3 enrollment in 544 schools, and CRLA test-takers exceed Grade 1 to 3 enrollment in 2,129 schools. This happens because the tests and the enrollment count fall at different points in the year.

Region boundaries and labels that changed

Four region changes need handling. ARMM became BARMM in SY 2019-20. Cotabato City moved from Region XII to BARMM in SY 2021-22. The Negros Island Region appeared in SY 2024-25 with 21 divisions from Regions VI and VII. Sulu moved from BARMM to Region IX in SY 2025-26.

Division names are not unique nationally. San Carlos City exists in Region I and in Negros Occidental, and San Fernando City exists in La Union and in Pampanga. Five divisions were renamed, among them Compostela Valley to Davao de Oro, Munoz Science City to Science City of Muñoz, and Taguig to Taguig City and Pateros. The series assign every school year to its division’s SY 2025-26 region, keyed on region and division together.

The sector and management labels changed too. “Local Government” management fell from 1,054 schools in SY 2018-19 to 189 in SY 2019-20. The drop happened because 858 schools were recoded to “DepEd”, not transferred. SY 2025-26 uses “SUCsLUCs” for “SUC/LUC”, “DepED Managed” for “DepEd”, and “SCHOOL ABROAD” for overseas schools. It also uses Yes/No for the True/False offering flags and labels with trailing spaces. The README documents none of these changes. The ELLNA and CRLA files spell the same languages differently (Ilokano/Iloko against Ilocano). All labels were mapped to one scheme.

Matching schools across files

The six SY 2023-24 files share the same set of 60,167 IDs, and the learner file repeats every enrollment column of the enrollment file exactly. Joins among them therefore lose nothing. The NAT file (6,636 schools) and the ELLNA file (5,752 schools) match the SY 2023-24 list completely, and they agree with it on region and division in every row. The CRLA file (6,121 schools) matches 6,090 of them; the other 31 schools opened in SY 2024-25.

Assessment coverage: five regions and 44 divisions

The assessment files cover CAR, Region I, Region III, Region VIII, and Region IX. They cover 44 of those regions’ 63 divisions, the same 44 for all three tests, and none of the other 13 regions. Inside the covered divisions, the ELLNA covers a median of 56% of public schools with Grade 3 learners, with a range from 5% to 100%. Twelve divisions are at or near complete coverage.

Covered and uncovered schools in those divisions are similar in size and annex status; the median Grade 3 enrollment is 25 against 23. They differ on the inputs by at most a quarter of a standard deviation. But the selection rule is not published anywhere in the files, and the outcomes of the uncovered schools are unknown. Every assessment finding in this series is therefore a statement about the assessed schools in the 44 covered divisions.

Region Division Public schools With Grade 3 learners ELLNA schools ELLNA coverage of Grade 3 schools NAT public schools CRLA schools
CAR Abra 317 261 174 67% 153 195
CAR Apayao 201 165 55 33% 64 72
CAR Baguio City 68 45 44 98% 44 42
CAR Benguet 428 362 19 5% 204 218
CAR Ifugao 266 225 164 73% 152 171
CAR Kalinga 194 160 57 36% 56 58
CAR Mt. Province 267 199 14 7% 14 14
Region I Candon City 34 29 25 86% 25 25
Region I Dagupan City 39 34 34 100% 34 34
Region I Ilocos Norte 364 314 92 29% 107 110
Region I Ilocos Sur 506 445 190 43% 184 195
Region I La Union 386 314 177 56% 157 175
Region I Laoag City 39 34 34 100% 34 34
Region I Pangasinan I, Lingayen 661 550 268 49% 266 260
Region I Pangasinan II, Binalonan 559 463 183 40% 184 185
Region I San Carlos City 76 60 59 98% 59 60
Region I Urdaneta City 64 46 45 98% 44 46
Region III Aurora 176 137 77 56% 74 77
Region III Bataan 214 168 69 41% 70 70
Region III Bulacan 540 441 176 40% 173 176
Region III City of San Jose Del Monte 58 37 37 100% 36 37
Region III Mabalacat City 55 40 40 100% 40 40
Region III Nueva Ecija 753 629 249 40% 246 250
Region III Pampanga 565 439 215 49% 215 217
Region III San Jose City 59 50 49 98% 48 49
Region III Science City of Muñoz 36 34 32 94% 32 32
Region III Tarlac 505 430 216 50% 207 219
Region III Tarlac City 98 88 88 100% 88 87
Region III Zambales 315 257 116 45% 115 117
Region IX Pagadian City 82 62 62 100% 62 62
Region IX Zamboanga Sibugay 507 412 155 38% 131 166
Region IX Zamboanga del Norte 759 647 290 45% 282 290
Region IX Zamboanga del Sur 807 674 285 42% 283 290
Region VIII Biliran 143 123 87 71% 87 87
Region VIII Borongan City 58 51 50 98% 49 51
Region VIII Calbayog City 174 156 156 100% 149 157
Region VIII Catbalogan City 58 50 50 100% 48 50
Region VIII Eastern Samar 466 412 176 43% 180 189
Region VIII Leyte 1,275 1,088 526 48% 516 546
Region VIII Northern Samar 607 520 332 64% 340 342
Region VIII Ormoc City 98 81 81 100% 80 81
Region VIII Samar (Western Samar) 738 663 324 49% 322 331
Region VIII Southern Leyte 343 294 134 46% 121 136
Region VIII Tacloban City 60 47 46 98% 47 47

The NAT adds a second limit. It tested a census of schools but only a sample of learners in each school, a median of six per school, and the sampling design is not described. The technical notes also caution that test difficulty differs from year to year, so this series compares none of the assessments across years.

What cannot be computed from the files

  • Seats per learner. The technical notes define a seat-learner ratio from kindergarten chairs, armchairs, and desks; the facilities file has no seat columns.
  • Gross and net enrollment rates. They need Philippine Statistics Authority (PSA) population projections, which are not in the release.
  • Shares of proficient learners. The NAT and ELLNA files give school means. The proficiency bands can be applied to a school’s mean, not to its individual learners.
  • Annex-adjusted ratios. The technical notes ask for annex teachers and learners to be added to the mother school, but no mother-annex mapping is published. Dropping annexes changes no result.
  • Trends in test scores. The technical notes rule them out, and the files hold one round of each test.
  • Anything about individual learners. Every file is a school aggregate.

What DepEd could change in future releases

  • Publish the assessment selection rule and the NAT learner-sampling design with each release.
  • Keep one coding scheme across years and keep zeros as zeros; document any change in the README.
  • Document the CRLA school-level aggregation and label its share columns as proportions.
  • Add seat counts to the facilities release and a mother-school ID for annexes.
  • Add range checks to the BEIS entry forms for ICT and personnel counts.
  • Publish a division-to-region history table so that boundary changes can be applied consistently.

Education Data, Transparency and Research covers what EDCOM II found wrong with DepEd’s data systems and what it recommends on releasing data.

How this audit was done

The original archives are never edited. Each file is profiled for its unit, rows, columns, types, key, and geography, and the data dictionary of all 2,216 columns follows the READMEs’ tables and naming conventions.

Each issue has a count and a recorded treatment, and the whole pipeline re-runs from the raw archives.

Frequently asked questions

Are DepEd’s public open datasets reliable?

Largely yes. The files have no duplicate school IDs, match cleanly across files, and hold few impossible values. There are four main cautions. The school modules cover public schools only, and the selection of assessed schools is undocumented. SY 2025-26 changed its conventions, and a few entry errors appear in ICT and personnel counts.

Why do DepEd’s NAT, ELLNA, and CRLA files cover only five regions?

DepEd released the NAT, ELLNA, and CRLA school-level files for a “selected set of schools” only. The selected schools are in CAR, Region I, Region III, Region VIII, and Region IX, in 44 of their 63 divisions, and DepEd has not published the selection rule.

Can DepEd’s assessment files rank individual schools?

Not reliably for the NAT, where each school mean rests on a median of six sampled learners. The defensible uses are division-level estimates with confidence intervals and school means shrunk toward their division mean.

Download the data

The cleaned, harmonized DepEd files behind this series are free to download in CSV and Parquet, with a codebook and checksums, under the terms of use set out in the release. Download Cleaned DepEd School Data, SY 2017-18 to 2025-26 documents every file and column. The files themselves are at data.nash.com.ph, starting with the README, the SY 2023-24 school master file (CSV, 8.0 MB) and the nine-year enrollment panel (Parquet, 17.2 MB).

More in this series

Guides for aspiring school heads

Analysis and charts by Mark Anthony Llego for NASH Reviewer. Data: Department of Education public datasets. Charts may be reused only with written permission from support@nash.com.ph.