Detention After ICE Arrests in Georgia

Limitaciones

Years are CALENDAR years of the arrest date (with a month column), not federal fiscal years. ICE and the Deportation Data Project report by fiscal year (October 1 to September 30), so a figure here will not match a published 'FY2026' number. Coverage is 2022-10-01 through 2026-08-06: calendar 2022 is October–December only and 2026 stops on 6 August, so both end years are partial and a trend line dips at each end for that reason alone. Scope, in DDP's own words: 'there are many detention stays that start without an ICE arrest (most often at the border, with a CBP arrest), and those detention stays are not included here.' This dataset describes what followed ICE arrests made in Georgia; it is not a count of people detained in Georgia. The arrest-to-stay match is approximate: DDP pairs each arrest with a detention stay whose book-in falls between five days before and ten days after the arrest, keeping the closest when several fit, and describes this as an imperfect method that works most of the time. detention_rate_after_arrest is therefore an ESTIMATE of the share of arrests that led to detention, not an exact administrative figure. This table is DDP-deduplicated to one arrest per 24-hour episode (duplicate_drop_row is false on every row; no further deduplication is applied). Its 27,567 Georgia arrests reconcile exactly to the companion dataset ice_arrests (28,220 Georgia rows in the same 2026-08-06 release, 27,567 after DDP's duplicate flag) — use ice_arrests for arrest counts and this dataset for the detention rate (23,594 of 27,567 Georgia arrests, 85.6%, led to a detention stay). Georgia is defined by DDP's apprehension_state_filled_in field (ICE's state back-filled from the event landmark; this table carries no separate original-state column). A further 2,018 arrests in ICE's Atlanta area of responsibility (Georgia, North and South Carolina) carry no state at all and are excluded from every state's count; an unknown share are Georgia arrests, so num_arrests is a lower bound, and their detention rate (55%) is well below Georgia's, so the published rate would be slightly lower if they could be placed. The source has no arrest-county field, so this dataset is STATE-ONLY: county_fips is NULL on every row and there is no county breakdown. State rows are a complete grid of every month × category combination, so a 0 count means no arrest matched that cell (zero is real; no suppression is applied at state grain), and detention_rate_after_arrest is NULL only where num_arrests is 0. Rates over a handful of arrests swing widely — check num_arrests before comparing cells. month = 'all' is the calendar-year total; every categorical's 'all' value is its total, so filter each to 'all' except the one you are breaking out, or the same arrest is counted on several rows. demographic is ICE's own gender determination (male/female), not self-reported; the 55 arrests recorded with gender Unknown count toward 'all' only, so male + female is slightly below all. citizenship_country names the 22 countries most frequent among Georgia records across the ICE and immigration-court sources (about 97% of Georgia arrests); every other country is 'other' and an unrecorded country is 'unknown'. Gender and citizenship are DDP's values on the joined arrest-stay record, which resolve more cases than the arrest file alone: this table records 160 fewer Unknown genders than ice_arrests, so a gender breakdown cell can differ from the same cell there by up to about 100 arrests (roughly 1 percent of a large cell), while citizenship cells differ by at most 3 arrests; every total, month, criminality and apprehension_method_group cell agrees exactly. criminality and apprehension_method_group are ICE's own classifications at the time of arrest, as recoded by DDP.