Try

US monthly employment and unemployment by occupation (CPS microdata)

Monthly employed persons, unemployed persons, unemployment rate and unweighted sample counts by detailed occupation (for example software developers, information security analysts, data scientists, engineers), by detailed and major occupation group, and for the whole civilian labor force, with age, sex, education and full-/part-time breakdowns for the whole labor force, every occupation group, and computer, mathematical, architecture and engineering occupations. Computed by FactIQ from the Current Population Survey (CPS) Basic Monthly person-level microdata (Census Bureau API) using the composited final weight PWCMPWGT, the same weight behind the BLS household-survey levels; the all-occupations employed series reproduces BLS series LNU02000000. Thousands of persons and percent, not seasonally adjusted, January 1994 to the latest published month. Detailed occupation codes changed in 2003, 2011 and 2020, so each code vintage is a separate series (ids occ1990_, occ2000_, occ2010_, occ2018_); cells with fewer than 30 sample persons are suppressed and months the Census Bureau did not publish (October 2025) are left missing. Not comparable with CES payroll employment.

CLASSIFICATION-BREAK NOTICE — CPS occupation series (cps_basic): the detailed occupation codes were redefined in January 2003, January 2011 and January 2020, so each code vintage is a SEPARATE series family with its own series_ids: occ1990_<code> covers 1994-01 to 2002-12, occ2000_<code> covers 2003-01 to 2010-12, occ2010_<code> covers 2011-01 to 2019-12, and occ2018_<code> covers 2020-01 onward. The same numeric code can mean a different occupation in another vintage (1020 is 'Computer software engineers' in occ2000_ and 'Software developers, applications and systems software' in occ2010_; software developers are 1021 in occ2018_). Occupation GROUP series (docc*, mocc*) changed once, in January 2003: docc1990_/mocc1990_ end in 2002-12 and docc2000_/mocc2000_ run from 2003-01 to the present. When a chart spans a break, keep each vintage as its OWN labelled column on a shared time axis — never chain vintages into one continuous line. Months the Census Bureau did not publish (October 2025) are missing, not zero, and must not be interpolated. These are not-seasonally-adjusted household-survey estimates: do not compare them with CES payroll employment, and average all twelve months before comparing with a BLS annual average.

In SQL, this dataset is the rows of census.series with dataset_code = 'cps_basic'.

Series

Showing 1–24 of 10,296