H-2B Non-Agricultural Labor Certifications

Limitaciones

Federal fiscal years (October 1 - September 30), not calendar years: `year` is the fiscal year in which DOL's Office of Foreign Labor Certification issued its final determination on the application (the decision date), the basis DOL itself reports on. DOL publishes no Georgia H-2B total: its selected statistics list only the top 10 states, and Georgia (2.1%-2.9% of national positions certified) is never among them. The national totals of the same disclosure files - positions requested, positions certified, and applications by outcome - match DOL's published H-2B Selected Statistics exactly in every year, FY2020-FY2026 (DOL's FY2023 PDF prints 15,502 applications processed, one more than its own outcome rows and the file hold, 15,501). FY2026 is PARTIAL: it holds decisions through June 30, 2026 only (DOL's cumulative Q3 file) and will be replaced when DOL posts the full year; do not compare it with complete years. No decision is dated before November 3, 2025 nationally or November 4, 2025 in Georgia, so October 2025 is empty - a gap that falls within the October 1 - November 12, 2025 federal government shutdown. Positions certified are job slots DOL approved for temporary foreign non-agricultural workers, not workers admitted or visas issued: the employer must still petition USCIS - where cap-subject workers count against the annual H-2B visa cap - and each worker must obtain a visa, so fewer may arrive. Each fiscal year counts the determinations issued in it, as in DOL's own figures. Withdrawn, denied and rejected applications certify no positions (DOL leaves the count blank, counted 0 here); rejected applications are incomplete filings returned without a decision on the merits, and 5 FY2023 rejected applications left the requested count blank, counted 0. Geography is the application's FIRST worksite (its worksite state, county and ZIP). 67% of the FY2020-FY2026 Georgia applications report work at additional worksites (DOL's Appendix A), which may lie in other counties or states; those sites are not used, so each application is counted once, where its first worksite is, and an application whose first worksite is outside Georgia is not counted even if it also lists Georgia sites. Employer, agent and attorney addresses are never used. Worksite counties are checked against the worksite street address. The Census Bureau's geocoder located the Georgia worksite addresses in 2020 Census geography and matched 1,447 applications (75%) to the same street address (house number and street, or intersection) in the same ZIP code. For 1,348 of them the address lies in DOL's county. For 88 applications (2,470 positions certified; FY2020 7, FY2021 11, FY2022 10, FY2023 10, FY2024 12, FY2025 22, FY2026 16) it lies in another county and a second sign agrees DOL's county is wrong - DOL's county holds under 5% of the worksite ZIP's residents, or the address lies at least 500 m from DOL's county, too far for a small placement error to explain, in a county where some of the ZIP's residents live - so the address's county is used. It also gives a county to 4 applications DOL left blank (0 positions certified). For 7 applications (59 positions certified) the address points to another county but lies within 500 m of DOL's county, where a geocoder can err, and DOL's county holds at least 5% of the ZIP's residents, so DOL's county is kept; some are DOL errors and some geocoder errors, and the geocoder alone cannot tell which. 18 applications (196 positions certified) are at addresses the geocoder cannot match (or matches in a county the records contradict) whose county the county's own tax-parcel and address-point records settle (17 at landmark sites whose county parcel is filed under another address and was found by owner name or map location, each confirmed by an independent source), and take that county: 0 (0 positions) move from the county DOL gave, 0 fill a county DOL left blank, and 18 confirm DOL's county. Where the address could not settle it, DOL's own worksite county is sometimes impossible for the worksite ZIP code - none of the ZIP's 2020 population lives in it. Such a county is replaced only when the right one is certain: at least 99% of the ZIP's population lives in one other county, and the worksite city is the ZIP's postal name or a Census place in that county, and not also a place or postal name in the stated county. 1 Georgia application (0 positions certified) was moved to the ZIP's county this way (FY2025 1); it sits in 1 ZIP code, 0 wholly inside the new county and 1 at least 99% inside it. Another 8 applications (216 positions certified) keep a stated county that is impossible for their ZIP, because the evidence falls short of certain: in 2 the ZIP straddles a county line and in 6 the city does not match the ZIP (the ZIP itself may be mistyped). Their county may be wrong. DOL left the worksite county blank on 17 Georgia applications (FY2022 2, FY2023 7, FY2025 7, FY2026 1). 4 of them are placed by their street address (above); 11 (48 positions certified) are placed in the county holding most of the population of their worksite ZIP code (2020 Census); 1 of them sits in a ZIP code that straddles a county line. Where DOL does name the county (as corrected above), the ZIP's dominant county disagrees with it on 127 of the 1,895 Georgia applications whose ZIP is in the crosswalk (6.7%), so a recovered county may be the neighboring one. 2 FY2022 denied applications could not be placed (worksite ZIP 30026, 3026, not a Georgia ZIP in the crosswalk) and count in the state row only, so that year's county rows sum to 2 applications, 6 positions requested and 0 positions certified fewer than the state row. In every other year county rows sum exactly to the state row. `cap_status` says whether an application's workers count against the statutory H-2B cap, and its source changes with DOL's form, so the series breaks in FY2024: through FY2023 it is the application's single yes/no cap-exempt flag, so every application is wholly 'cap_subject' or 'cap_exempt' and none can be 'mixed'; from FY2024 the form reports how many of the requested workers are subject to the cap and how many are exempt, and an application with some of each is 'mixed' (FY2024 also includes 24 Georgia applications filed on the older form, classified by their flag); 'unknown' is a FY2026 rejected application that reports neither. It is an application-level label: the positions of a 'mixed' application are not split, because DOL reports the subject/exempt split for requested workers only, not for certified ones. `case_status` is the final determination. Certifications DOL marks '(Returned)' count as certified or partially certified, as in DOL's own totals, but the employer returned them to DOL: DOL's H-2B guidance lets an employer return a certification it no longer needs - typically unused, most often because the H-2B visa cap was reached - and file the same job again with a later start date (DOL does not define the label in the record layout; the return procedure is in its H-2B FAQs, https://www.dol.gov/sites/dolgov/files/ETA/oflc/pdfs/H-2B_Visa-Cap_FAQs_03.06.2020.pdf, and its H-2B webinars say only an unused certification may be returned). So the same job slots can be certified twice. Georgia's returned certifications: FY2020 14 applications, 399 positions (10%); FY2022 5 applications, 52 positions (1%); FY2023 6 applications, 195 positions (3%); FY2024 7 applications, 88 positions (1%); FY2025 6 applications, 121 positions (2%); FY2026 37 applications, 1,198 positions (20%). DOL's FY2021 file carries no '(Returned)' status at all, nationally, so returns cannot be identified in that year. `soc_occupation` names the five occupations (SOC codes) with the most Georgia positions certified over FY2020-FY2026 and groups every other occupation as 'other'. `nature_of_need` is the application's stated basis for the temporary need (one-time occurrence, seasonal, peakload, intermittent), as the employer filed it. County rows are totals only (every categorical 'all'); the breakdowns are statewide. No suppression: rows are applications (job orders), not people, so the domain's county small-cell rule does not apply, small county counts are published, and every zero is a real zero. Wages and employer identity are not published here.