Linked ASEC Notes

Back to the Samples Page

Linked ASEC files are built from the cross-sectional Annual Social and Economic (ASEC) supplement to the Current Population Survey (CPS). This document describes how the linked files are constructed by IPUMS CPS and how variables in linked files are named.

LNKFW1YWT should be used as the weight for person analyses.

Linked ASEC files can be found by selecting the "Longitudinal, 1 Year Apart" radio button on the IPUMS CPS sample selection page. Linked ASEC files deliver files with person records that have been matched across adjacent years of the ASEC. For example, records in the 2017-2018 linked ASEC file include all individuals who are in both the 2017 and 2018 ASEC files. Individuals who are only in the 2017 or 2018 ASEC data are excluded from the linked ASEC file. Individuals who belong to the ASEC oversample (see ASECOVERP) are also excluded from the linked ASEC file.

Each record in the linked file represents a pair of ASEC records matched using CPSIDP. Variables are renamed to include an _1 or _2, corresponding to the respondent's first or second ASEC observation, respectively. For example, while AGE is available in the 2017 and 2018 cross-sectional files, AGE is renamed to AGE_1 and AGE_2 in the linked ASEC file, which correspond to AGE in 2017 and AGE in 2018, respectively. In instances where a variable is only available in one year of the linked ASEC file, the variable will only have data for that time point. For example, SPMCAPHOUS is available beginning in 2010. In a longitudinal extract including the 2009-2010 ASEC samples, SPMCAPHOUS_1 (the 2009 data) will have no values, however, SPMCAPHOUS_2 (the 2010 data) will.

Currently, IPUMS CPS provides linked ASEC data from 1978 forward. Note that a pair of linked samples only becomes available for selection via IPUMS CPS when the second year in the pair has been released. Some pairs of years cannot be linked. The 1976, 1977, 1985, and 1995 ASEC files cannot be linked to the ASEC file from the following year due to incompabibility of Census Bureau identifiers across these pairs of years (Flood et al., 2020; Drew et al., 2014).

Back to Top