Context accompanying the corrected census
dfe_education_context.csv contains all 222 official observations for England state-funded and special-school pupils aged 15 in 2020/21, followed into higher education by age 19 in 2024/25. Every recorded sex, FSM status and major/detailed ethnicity category is retained, including the recorded Unknown-sex observation. Original published percentages, pupil denominators and HE entrant counts remain intact. Source: DfE release, 9 July 2026; dataset.
dfe_education_comparisons.csv pivots these observations into the paper's complete 24-row table: 19 detailed categories, four published major-group subtotals and the national total. It reports Male/Female x FSM/non-FSM rates and their denominators/entrant counts, all-sex totals and the original category labels. Subtotals overlap their detailed components. Non-FSM is the source's All Other Pupils category; FSM is eligibility at age 15, an administrative low-income indicator rather than the whole working class. The paper discusses the official education context separately from its model-builder classifications.
dfe_education_source.json records the original DfE download, archive member, source SHA-256, licence and exact cohort filter. python scripts/build_education_context.py reproduces the comparison CSV and paper table from the released 222-row extract without a network request.
census_context.json retains the already-fixed hardware price bands, AI-centre city list, local-authority deprivation rule and schooling definition. uk_hardware_prices.csv preserves dated UK listings; prices describe comparable accelerators, not historical payments or whole-system cost. england_imd2025_local_authorities.csv contains the official local-authority evidence used for the regional descriptions. London without a borough remains unresolved for local-authority deprivation.
builder_descriptive_summary.csv in the data root reports person-level route, schooling, provenance, compute, price-band, AI-centre and deprivation totals. Multiple-response categories may overlap; the denominator is explicit. Unknowns stay separate. None of these fields is a name-based demographic classification.
Official statistics retain Crown copyright and their publishers' Open Government Licence terms. Source links are preserved in the data.