US monthly employment, income, SNAP and job-loss outcomes from SIPP microdata

Monthly employment share, unemployment share, not-in-labour-force share, unemployment rate and employed persons by age group, sex and education (bachelor's degree or more vs less), median monthly person income, earnings and household income, SNAP receipt rate, unemployment-compensation receipt rate and monthly poverty rate (household income below the monthly poverty threshold), employed persons in computer, mathematical and engineering occupations, and longitudinal job-loss outcomes: job-loss rate, share of job losers with a job again 3, 6 and 12 months later, SNAP receipt 6 months after a job loss and the median change in household income 6 months after a job loss (monthly, and quarterly-pooled versions with the _q suffix, e.g. sipp.reemp_6m_all_q, that pool the job losses of a calendar quarter and are dated the first day of the quarter). Computed by FactIQ from the Survey of Income and Program Participation (SIPP) public-use person-month files with the monthly person weight WPFINWGT: the 1996, 2001, 2004 and 2008 panel core wave files and the 2014+ redesign panels (Census Bureau microdata API for 2022+ data years, public-use CSV zips before). Thousands of persons, percent and nominal monthly dollars; not seasonally adjusted; March 1996 to December 2023 with three gaps where no panel has its full sample in the field: 1999-12 to 2000-12, 2003-10 to 2003-12 and 2007-10 to 2008-07. Every month comes from exactly one panel (the newest panel whose full sample covers the month: 1996 panel 1996-03 to 1999-11, 2001 panel 2001-01 to 2003-09, 2004 panel 2004-01 to 2007-09, 2008 panel 2008-08 to 2012-12, 2014+ panels 2013-01 onward); panels are never averaged, so a small level shift at each panel change is a sample/questionnaire change, not an economic event. Each 2014+ data year covers the previous calendar year and is released once, 12-18 months after the reference year. Unweighted sample counts (n_ series) accompany every estimate. Key series: sipp.emp_share_all, sipp.urate_all, sipp.medinc_hh_all, sipp.snap_rate_all, sipp.poverty_rate_monthly_all, sipp.emp_occ2018_1021 (software developers), sipp.jobloss_rate_all, sipp.reemp_6m_all, sipp.snap_after_loss_6m_all, sipp.hh_income_drop_6m_all, sipp.jobloss_rate_all_q. Occupation series carry the code list in force: emp_occ1990_ (1996 and 2001 panels), emp_occ2000_ (2004 and 2008 panels), emp_occ2010_ (2014 panel), emp_occ2018_ (2018+ panels); they are separate series and never spliced. SIPP collects each month retrospectively at an interview, so employed persons run about 2-3 percent above cps_basic.emp_all; in the 1996-2008 panels monthly job-loss rates spike every fourth month (the month between two interviews), so prefer the quarterly _q series there; follow-up series end 3, 6 or 12 months before the latest month because censored events are excluded.

CLASSIFICATION-BREAK NOTICE — SIPP occupation series (sipp): job 1 occupation is coded with the Census occupation code list in force for each panel: 1990 codes in the 1996 and 2001 panels (series emp_occ1990_<code>, three-digit codes, e.g. emp_occ1990_064 = computer systems analysts and scientists), 2000 codes in the 2004 and 2008 panels (emp_occ2000_<code>, e.g. emp_occ2000_1010 = computer programmers), 2010 codes in data years 2014-2017 (reference months 2013-01 to 2016-12; emp_occ2010_<code>, e.g. emp_occ2010_1020 = software developers, applications and systems software) and 2018 codes in data years 2018 onward (reference months 2017-01 onward; emp_occ2018_<code>, e.g. emp_occ2018_1021 = software developers). The same numeric code can mean a different occupation in another vintage, so each vintage is a SEPARATE series and must be kept as its own labelled column on a shared time axis — never chain them into one continuous line. PANEL-BREAK NOTICE — every sipp month comes from exactly one SIPP panel, the newest panel whose full sample covers that month (1996 panel: 1996-03 to 1999-11; 2001 panel: 2001-01 to 2003-09; 2004 panel: 2004-01 to 2007-09; 2008 panel: 2008-08 to 2012-12; 2014+ panels: 2013-01 onward). Panels are never averaged, so each panel change is a sample and questionnaire break (also between the 2014 panel and the 2018+ panels at 2016-12/2017-01); the months 1999-12 to 2000-12, 2003-10 to 2003-12 and 2007-10 to 2008-07 have no value because no panel had its full sample in the field. Follow-up series (reemp_3m/6m/12m, snap_after_loss_6m, hh_income_drop_6m and their quarterly _q versions) end 3, 6 or 12 months before the last cross-sectional month because censored events are excluded, not because data are missing. In the 1996-2008 panels the monthly job-loss rate spikes every fourth month (the month between two interviews); the quarterly _q series pool three months and do not show this pattern.

In SQL, this dataset is the rows of census.series with dataset_code = 'sipp'.Copy SQL to list series