Skip to content

March (ASEC) extracts

The EPI CPS March extracts cover the Annual Social and Economic Supplement (ASEC) to the Current Population Survey from 1962 to 2026. Files are named epi_cpsmarch_{year}.

Years Source
1962 - 1997 Unicon CPS March
1998 - 2026 Census CPS ASEC public use files

Version 2026.10.13 rebuilds the March (ASEC) extracts for all years and adds 2019–2026. Results from earlier versions of the March extracts will not always match the current version. This page describes what changed, the breaks in the underlying survey, where the extracts differ from IPUMS, and which variables are limited to certain years.

What changed from previous versions

Renamed and removed variables

Previous name Current name Notes
faminc (Basic Monthly/ORG) i_faminc Intervalled family income; faminc is now continuous family income in the March extracts
faminc_det i_faminc_det Intervalled family income, detailed

New variables

Weights

  • asecwgt, hhwgt, and famwgt use the same population controls as Census's published historical income tables:
    • 1980 and 1993: Unicon alternate weights, based on the 1980 and 1990 Censuses
    • 2011: weights re-estimated by Census using 2010 Census-based controls
    • 2021: weights re-estimated by Census using 2020 Census-based controls
    • 2025: Vintage 2025 revised weights
  • hhwgt is missing for group quarters units, which Census does not count as households. In 1962–1976, hhwgt is the asecwgt of the household head. In 1968–1975, households with no _relhd head take the weight of the head identified by hhrel instead of being missing.
  • famwgt for 1963, 1964, and 1966 is the asecwgt of the family reference person, so it carries the same adjustment factors as asecwgt. Zero and negative famwgt values are missing.
  • asecwgt for 1963 and 1966 is adjusted using the IPUMS methodology.

Survey and series breaks

  • 2014 split-sample redesign. The 2014 ASEC was fielded with two questionnaires: 3/8 of the sample got the redesigned income and health insurance questions and 5/8 got the traditional questions. redesign identifies which group a record belongs to. Both groups are included, and their weights are scaled by 3/8 and 5/8 so the combined file sums to the population. When you analyze one group alone, divide its weights by 3/8 (redesigned) or 5/8 (traditional) to get population totals.
  • 2019 processing-system change. Starting with the 2019 ASEC, Census processes the redesigned questionnaire with an updated system. Several income, poverty, and health insurance variables come from different source variables before and after 2019. For health insurance, the 2018 bridge file (2018 respondents run through the updated system) provides an overlap year: see hicovly4 (legacy processing) and hicovly5 (updated processing).
  • Health insurance coverage series. There is no single consistent series for coverage last year. Use hicovly1 (1988–2000), hicovly2 (2001–2010, Census's 2007 revision), hicovly3 (2000–2014, traditional questions), hicovly4 (2014–2018, redesigned questions, legacy processing), or hicovly5 (2018–present, updated processing). The 2001 extract uses the expanded SCHIP sample file.
  • Population control revisions. Years in which weights move to new population controls (1980, 1993, 2011, 2021, 2025) can produce level shifts in weighted counts. See Weights.
  • Unicon to Census source files (1998). Variables come from Unicon files through 1997 and from Census public use files starting in 1998. Source variables for each era are listed on each variable's page.

Differences from IPUMS

Variable EPI IPUMS Years
asecwgt Alternate weights in 1980 and 1993, revised weights in 2011, 2021, and 2025, matching Census's historical income tables ASECWT uses original weights in 1980, 1993, and 2025, and revised weights in 2010 and 2020 1980, 1993, 2010–2011, 2020–2021, 2025
offpov Agrees with Census famlis and IPUMS POVERTY OFFPOV uses related subfamilies' own thresholds (2000) and codes a zero-income unrelated subfamily without a threshold as above poverty (2001) 2000–2001
hiprivly Adds children covered by private plans (chhi), consistent with hicovly1 COVERPI excludes them 1988–1994
ftptstat Not in labor force is coded 1 WKSTAT puts it in NIU All years

Coverage gaps by year

Many March variables start after 1962. Some notable gaps:

  • Geography: statefips begins in 1963 and countyfips in 1996. cbsafips and cbsasize begin in 2005.
  • Weights: famwgt begins in 1963.
  • Race and ethnicity: hispanic begins in 1971, wbhao in 1988, and wbhom in 2003. In 1971–2002, "don't know" and "NA" responses to Hispanic origin are missing in hispanic but counted as not Hispanic in wbho and the other race variables.
  • Poverty: official poverty variables begin in 1968 and SPM variables in 2010. Universe definitions change in 1980 (see offpov).
  • Income: i_faminc is not available in 1989.
  • Health insurance: coverage last year begins in 1988 (calendar year 1987), and current coverage (hicov) in 2014. Employer-plan variables hiownemp and hiownpaid begin in 2019.
  • Citizenship and labor force detail: citizen, discwork, and dlfstat begin in 1994.

See each variable's page for full availability by year.