Dataset
School enrolment, population and learning assessment — 2023 and 2024
2,471 enrolment records across two school years, the school-age population that is their denominator, and 2,798 learning assessment results in two rounds — the three files gross and net enrolment, cohort survival and learning outcomes each need.
syntheticEducationStudent enrolmentStudent attendanceLearning outcomesDropout analysisSchool feeding
- Rows
- 5,349
- Variables
- 18
- Period
- School years 2023 and 2024
- Licence
- CC BY 4.0
- Completeness
- 99%
Standards and methodologies
Sustainable Development Goals (SDG)UNICEF indicator definitionsMultiple Indicator Cluster Survey (MICS)
Files
Files are versioned by filename. A corrected release ships as .v2.csv rather than replacing the file in place, so an analysis pinned to v1 keeps reproducing.
Data dictionary
| Variable | Type | Description | Allowed values |
|---|---|---|---|
student_id | string | Pseudonymous student identifier. Shares its values with the school roster and attendance files, so the three join. | — |
school_id | string | School the student was enrolled at. A student who transferred appears under both schools in the same year. | — |
admin2 | categorical | District the school sits in. | Nord-Ouest, Artibonite, Centre, Sud |
school_year | integer | School year the row describes. 2023 is complete; 2024 is still running, so every 2024 row carries the status enrolled. | — |
grade | integer | Grade the student was enrolled in. Grades 1 to 6 are primary; the official age for grade g is g plus five. | — |
age_yearsyears | integer | Age in completed years at enrolment. One school records part of its intake in months. | — |
sex | categorical | Sex of the student. | f, m |
disability_reported | boolean | Whether a disability was reported at enrolment. Self-reported and under-reported, as everywhere. | true, false |
displacement_status | categorical | Displacement situation of the household. Blank for a block of students, and not missing at random. | resident, idp, returnee, host |
repeating | boolean | Whether the student was repeating the grade. Recorded unreliably in one district. | true, false |
end_of_year_status | categorical | What happened at the end of the school year. The field a cohort analysis is built from. | promoted, repeated, dropped-out, transferred-out, completed-final-grade, enrolled |
reference_year | integer | Population file: the year the projection refers to. | — |
projected_population | integer | Population file: projected children of that single year of age in that district. A projection from a 2015 census compounded at 2.1% a year, not a count. | — |
assessment_round | categorical | Assessment file: which round the result comes from. The two rounds do not have the same number of items. | baseline, endline |
domain | categorical | Assessment file: what was assessed. | literacy, numeracy |
items | integer | Assessment file: number of items in that round's instrument. 40 at baseline and 50 at endline, which is why raw scores are not comparable across rounds. | — |
raw_score | integer | Assessment file: items answered correctly. Compare as a proportion of items, or not at all. | — |
proficiency_band | categorical | Assessment file: the equated band. This is the measure that carries across rounds; the raw score is not. | below-minimum, minimum, proficient, advanced |
Provenance
Data quality
- Fully generated. No real student is described and no school code refers to a real school.
- Gross and net enrolment are 109.2% and 97.4%. 36.4% of primary enrolment is above the official age for its grade, and 10.8% of primary enrolees are twelve or older. A dashboard quoting one ratio without saying which is the commonest error in this sector, and a gross ratio above 100% is not an error at all.
- The population file is a projection, not a count: a 2015 census compounded at 2.1% a year. Nine years of compounding is a factor of 1.21, so about a fifth of the denominator is an assumption — the same structure as the immunisation denominator in the public health course.
- The two assessment rounds used 40 and 50 items. A raw literacy score of 30.9 at endline and 21.2 at baseline is 61.8% against 53.1%, and the difference between those two framings is the whole of the comparability problem. Only the proficiency bands were equated.
- The assessment subsample is not random. 741 students sat the baseline and 658 the endline, and the ones who did not sit it are the ones attending least. Across all comers literacy rises from 53.1% to 61.8% of items; across the 585 students who sat both rounds it rises from 56.8% to 63.7%. About 1.8 of those 8.7 points is who turned up rather than what they learned.
- Reference figures for a correct analysis: 2023 end-of-year status is 70.4% promoted, 9.8% repeated, 10.2% dropped out, 5.7% completed the final grade and 2.9% transferred out; dropout is 19.5% among over-age students against 4.8% among in-age.
Known issues
- SCH04 records part of its intake's age in months rather than years, producing twenty-one ages above 20. They are impossible against a primary register and will not be caught by a mean, only by a range check.
- Centre records repetition unreliably: about 38% of its genuine repeaters are flagged as new entrants. Its repetition rate reads 7.4% against 12.0% to 12.3% elsewhere while its over-age share is the highest of the four districts, and that combination is a recording habit rather than a school effect.
- Displacement status is blank for 42 students, and the blanks are concentrated among displaced and returnee households — the field an enumerator skips when a family has just arrived. Dropping the blanks removes the group with the highest dropout.
- Twelve students appear twice in the same school year, once under each school, because a transfer was recorded as a new enrolment rather than a move. Any enrolment count that does not deduplicate on student and year is twelve too high.
Worked examples
Python
Gross and net enrolment from the same register
Builds both ratios against the projected denominator, decomposes the gap into over-age enrolment, and shows what the projection contributes to a figure above 100%.
Described here; the notebook is still being written.
Two assessment rounds that are not comparable
Contrasts raw scores, proportions of items and equated bands across the two rounds, and estimates the change on the students who sat both.
Described here; the notebook is still being written.
R
A cohort through the school year
Follows the 2023 cohort through promotion, repetition and dropout, and disaggregates dropout by over-age status, disability and displacement.
Described here; the notebook is still being written.