Per-Pupil Expenditure (FESR)

Limitations

Two detail levels (district, school) in one fact table, split by parquet filename; school_code is NULL on district rows. Detail level is read from the geography-key NULL pattern. The district level covers FY2014-FY2025 (180 traditional systems, codes 601-793, no charters/RESAs/state agencies); the school level covers FY2017-FY2025 (all schools, including charter authorizers). School rows for each fiscal year are the schools listed by that fiscal year's OWN FESR publication; later rolling-window files that republish the year only fill NULL cells on those schools, and school codes that appear only in a later file's back-window (GOSA re-coded schools restated under new codes — 11 FY2017 and 5 FY2018 keys, including the four FY2017/FY2018 per-pupil-only rows the 2019 file annotates as involved in a FY18 code change) are dropped and ledgered in _transform_manifest.json as school_code_restated_in_later_publication, so a district-year never carries two breakouts of the same money; a quality check keeps 3-digit-district school fte_enrollment sums within 5% of the district row. Verified in gold: in every 3-digit district-year the school rows sum to the district row's fte_enrollment exactly, and to its total_expenditures within 0.3% in every district-year except two in FY2025 (district 705 +1.85%, district 755 -0.88%); the two levels come from independent GOSA downloads, so exact dollar agreement is not guaranteed. Level-specific columns are NULL on the other level: federal_expenditures / state_local_expenditures are district-only; included_* / excluded_expenditures / total_federal_expenditures / total_state_local_expenditures / pct_of_district_enrollment are school-only. The district federal_expenditures and the school total_federal_expenditures are the same federal share of total_expenditures but are kept as separate columns (different source naming; schools also publish the distinct pre-allocation included_* split). District federal/state-local breakdowns are NULL for FY2014 (the only source covering FY14 published totals only). The 2025 district workbook adds a district excluded-expenditure figure (District_Excluded_Amt_23/_24/_25, FY2023-FY2025) that is NOT carried: it is not the district total of the school excluded_expenditures (in FY2025 the district's schools' excluded amounts sum to a median ~7% of it), so it does not fit that school-only column; the omission is ledgered in _transform_manifest.json read_loss. FY2022 methodology change: from FY22 school total_expenditures (and per_pupil_expenditure) additionally include Allocated CCA and Alternative Program costs (still so in FY2025, per that file's variable key), so FY22+ school totals are not strictly comparable to FY17-FY21. This is the FESR included-cost per-pupil series — totals intentionally differ from the k12_expenditures function-level expenditure ledger. Placeholder $0 school expenditure amounts on unreported/non-compliant FY17-FY18 rows are NULLed per §4b (see _transform_manifest.json masked_values). One FY2021 school row (7830629/0629) publishes the literal 'Non-Compliant' publisher status in all three per-pupil columns; those three values are NULL here (recorded in masked_values; the status is preserved as is_non_compliant in the financial_efficiency_star_rating topic). Negative dollar values occur in three places and are faithful GOSA restatements preserved as-is: per_pupil_expenditure / total_expenditures / included_expenditures and their state-local components on three district 660 (Fulton County) school rows in FY21-FY22; excluded_expenditures on a few rows every year (52 rows, FY2017-FY2025); and included_federal_expenditures on 24 rows across nine districts (FY2017-FY2025). Currency is nominal dollars (not inflation-adjusted). No cell-level small-cell suppression exists in this source; NULLs are structural or §4b masks, never suppression.