Retained (Held-Back) K-12 Students

Students held back a grade in Georgia public schools, with counts and the race/gender makeup of the retained group, by school and district, 2004-2024.

Source: Governor's Office of Student Achievement (GOSA)
Year coverage: 2004–2024
Updated: annual
Version: 1.0.0
Schema hash: 9a026e84857d

Detail levels available

StateDistrictSchool · default

Overview

Georgia Office of Student Achievement (GOSA) Retained K-12 dataset. For every Georgia public school, school district, and the state as a whole, reports the total number of students retained (held back) in grades K-12 during the school year, the total K-12 enrollment (2011 onward only), and a demographic breakdown of the retained population — counts and shares of retained by race and gender. The race axis carries six canonical codes (`native_american`, `asian_pacific_islander` — the bronze `Asian`/`Asians` label is a combined Asian/Pacific Islander bucket, with no separate Pacific Islander row — `black`, `hispanic`, `multiracial`, `white`); gender is `male`/`female`. Coverage spans the 2003-04 school year (filename year 2004) through the 2023-24 school year (filename year 2024) — 21 files.

Row grain: One row per year, district_code, school_code, demographic.

Key metric

pct_of_retained_cohort — Percent of Retained Cohort (proportion)

The single headline value most answers want.

JSON schema →Data contract (YAML) →

Variables

Every column in the fact table. The code is the exact field name the API returns.

Column	Type	Unit	Range	Null?	Description
`year`Year	integer	—	—	no	Ending calendar year of the school year (2024 = 2023-2024). Sourced from the bronze `School Year` / `School_Year` column for 2011-2024 (cross-checked against the filename); from the filename for 2004-2010 (no year column in those eras). Full description▾ Ending calendar year of the school year (2024 = 2023-2024). Sourced from the bronze `School Year` / `School_Year` column for 2011-2024 (cross-checked against the filename); from the filename for 2004-2010 (no year column in those eras).
`district_code`District Code	string	—	—	yes	3-digit GOSA district code (zero-padded) for standard districts; 7-digit code for state charter / state school networks. NULL for state-level aggregate rows. FK to districts dimension. Full description▾ 3-digit GOSA district code (zero-padded) for standard districts; 7-digit code for state charter / state school networks. NULL for state-level aggregate rows. FK to districts dimension. Joins districts; adds its label columns.
`school_code`School Code	string	—	—	yes	4-digit GOSA school code (zero-padded). NULL for district- and state-level aggregate rows. FK to schools dimension (composite key with district_code). Full description▾ 4-digit GOSA school code (zero-padded). NULL for district- and state-level aggregate rows. FK to schools dimension (composite key with district_code). Joins schools; adds its label columns.
`demographic`Demographic	string	—	—	no	Student subgroup the row covers (race or gender); 'all' is the whole retained group. Race uses the combined asian_pacific_islander bucket. Full description▾ Canonical demographic code (FK to demographics dimension). `all` is the overall retained cohort; race uses the combined `asian_pacific_islander` bucket (pre-1997 OMB convention: the six race-bucket retained counts sum exactly to the total retained at the state row in every year, so Pacific Islanders are folded in, never published separately); gender is `male`/`female`. Race values are mutually exclusive with each other, as are gender values; `all` overlaps both axes. 9 allowed values▾ `all` `asian_pacific_islander` `black` `female` `hispanic` `male` `multiracial` `native_american` `white` Joins demographics; adds its label columns.
`num_retained`Number Retained	integer	count	—	yes	Number of K-12 students in this demographic at this entity who were retained (held back) during the school year. For `demographic = all`, the total retained count. NULL when suppressed. Full description▾ Number of K-12 students in this demographic at this entity who were retained (held back) during the school year. For `demographic = all`, the total retained count. NULL when suppressed.
`num_students`Number Students	integer	count	—	yes	Total K-12 enrollment at this entity. Populated only on `demographic = all` rows in 2011 and 2013-2024. NULL for every non-`all` row (bronze publishes no per-demographic enrollment) and for every row in 2004-2010 (the wide-era bronze publishes no enrollment at all — its total columns are retained counts). 2012 is fully NULL: the 2012 bronze `Total Enrolled` is corrupt (inflated ~17x at every detail level, state row 27,864,309 vs ~1.65M expected) and is NULLed per the known-source-defect rule; 2012 `num_retained` is internally consistent and preserved. With `num_retained` on `all` rows this yields an overall retention rate. Full description▾ Total K-12 enrollment at this entity. Populated only on `demographic = all` rows in 2011 and 2013-2024. NULL for every non-`all` row (bronze publishes no per-demographic enrollment) and for every row in 2004-2010 (the wide-era bronze publishes no enrollment at all — its total columns are retained counts). 2012 is fully NULL: the 2012 bronze `Total Enrolled` is corrupt (inflated ~17x at every detail level, state row 27,864,309 vs ~1.65M expected) and is NULLed per the known-source-defect rule; 2012 `num_retained` is internally consistent and preserved. With `num_retained` on `all` rows this yields an overall retention rate.
`pct_of_retained_cohort`Percent of Retained Cohort	number	proportion	—	yes	Share of this entity's retained students who are in this demographic, on a 0-1 scale; a makeup share, NOT a retention rate, and NULL on the 'all' row. Full description▾ Share of the entity's total retained cohort who belong to this demographic, 0-1 scale (bronze 0-100 divided by 100; whole-percent precision in 2011-2024). NOT a within-demographic retention rate. In 2009 the published White and Multiracial percent columns are corrupt (both repeat the Male share); for those two demographics in 2009 the share is derived as num_retained / total retained from the internally consistent counts instead.

Filters

Query parameters the fact endpoint accepts. Comma-separated values are OR within a parameter; multiple parameters AND together.

Parameter	Kind	Allowed values	Notes
`year`	year_exact	any value	—
`year_min`	year_range	any value	—
`year_max`	year_range	any value	—
`detail`	detail	`districtsschoolsstates`	enum-enforced, default: schools
`district_code`	foreign_key	any value	multi-value
`school_code`	foreign_key	any value	multi-value
`demographic`	foreign_key	9 allowed values▾ `all` `asian_pacific_islander` `black` `female` `hispanic` `male` `multiracial` `native_american` `white`	multi-value, enum-enforced
`district_type`	dimension_attribute	7 allowed values▾ `commission_charter` `resa` `standard` `state_agency` `state_charter` `state_school` `state_special`	multi-value, enum-enforced
`demographic_category`	dimension_attribute	12 allowed values▾ `aggregate` `disability` `economic_status` `esol` `foster_care` `gender` `grade` `homeless_status` `migrant_status` `military` `race` `sped`	multi-value, enum-enforced

Example requests

Run these against the base URL — they return live data.

All rows (first page), schools level

curl "https://georgiacivicdata.org/api/v1/education/retained_students"

schools data for 2024

curl "https://georgiacivicdata.org/api/v1/education/retained_students?year=2024"

One district (code 601), 2024

curl "https://georgiacivicdata.org/api/v1/education/retained_students?district_code=601&year=2024&detail=districts"

Download 2024 as CSV

curl "https://georgiacivicdata.org/api/v1/education/retained_students?year=2024&format=csv"

Notes & limitations

NULL handling

Suppressed cells (too few students to report) are NULL, never zero.
Zero is a real, reported value.

Limitations

`num_students` (total K-12 enrollment) is available only on `demographic = all` rows for 2011 and 2013-2024. It is NULL for every non-`all` demographic row (bronze has no per-demographic enrollment), for every row in 2004-2010 (the wide-format bronze publishes no enrollment column), and for all of 2012 (corrupt bronze `Total Enrolled`, inflated ~17x, is NULLed); 8 2016 `all` rows are also NULL where source enrollment was unpublished. An overall retention rate (`num_retained` / `num_students`) is therefore computable only on `demographic = all` rows in 2011 and 2013-2024. `pct_of_retained_cohort` is a SHARE of the total retained cohort that belongs to each demographic (it sums toward 1.0 across race demographics and across gender demographics within an entity), NOT a within-demographic retention rate; it is NULL on `demographic = all` rows (100% by definition). Suppressed cells (cohort below the privacy threshold) null `num_retained` and `pct_of_retained_cohort`.