Repeated Sample (Cross-sectional and Longitudinal)

CPS
1

I plan to utilize both cross-sectional and longitudinal data from the Current Population Survey (CPS).

Before I proceed with my analysis, I have a few questions regarding the CPS datasets:

  1. Can I combine different months and years of the cross-sectional datasets to increase my analysis sample size? Are some of the samples the same individuals (repeatedly) over months or even over the years?
  2. Is the data in the Annual Social and Economic Supplement (ASEC) duplicating the same individuals (repeatedly) to the core dataset?
  3. Are there any supplemental data on volunteerism in the longitudinal data?
  4. Can I compare the same individuals over the years in the longitudinal data? Is there a variable to identify the same individuals in 2021-2022 and compare them to 2020-2021 to test the changes between 2020 and 2022?
  5. How was the longitudinal data collected?
2

Please see my answers below in line with your questions:

  1. Can I combine different months and years of the cross-sectional datasets to increase my analysis sample size? Are some of the samples the same individuals (repeatedly) over months or even over the years?

Yes you can combine multiple months and years. The same individuals are sampled repeatedly, as you can read about on this page about the CPS rotation pattern. This can be problematic in pooled samples when calculating standard errors because disturbances will be correlated across multiple observations of a single individual. The proper way to deal with this is not 100% clear. You may want to either keep only one observation per person (using a combination of CPSIDP and MISH), or cluster standard errors at the person level.

  1. Is the data in the Annual Social and Economic Supplement (ASEC) duplicating the same individuals (repeatedly) to the core dataset?

Yes, each year half of the ASEC sample is new, and half participated in the previous year’s ASEC. You can identify whether it is a household’s first or second year in the ASEC using a combination of YEAR and CPSID. Households will never appear more than twice in the ASEC. Please note that you can make a pre-linked longitudinal extract of the ASEC using the IPUMS extract system. To do this, select the “longitudinal, 1 year apart” option on the sample selection screen.

  1. Are there any supplemental data on volunteerism in the longitudinal data?

I will refer you to our ROPES tool to see which samples can be linked to a sample with a Volunteer supplement. I don’t believe anyone will appear in two Volunteer supplements, because they are two years apart. The maximum time between a household’s first and last appearances in the CPS sample is 16 months.

  1. Can I compare the same individuals over the years in the longitudinal data? Is there a variable to identify the same individuals in 2021-2022 and compare them to 2020-2021 to test the changes between 2020 and 2022?

See my answer to question 2

  1. How was the longitudinal data collected?

The longitudinal component of the CPS is based on repeatedly visiting or calling the residents of a given housing unit. An important thing to be aware of is that it is the housing unit (i.e. the building) that is sampled. If the family that is residing in the household moves away, they are not followed. Instead, the new residents are interviewed. You can read more about the details in the CPS technical documentation.