Mismatches after Linking ATUS to CPS Monthly Basic (MIS8)

TIME USE
1

Hi IPUMS Team,

I read the post about merging CPS and ATUS but encountered a huge sample loss while linking them, and I wonder why.

My research needs the teleworking variables from CPS (i.e., TELWRKDIFFCVD, TELWRKPAY, TELWRKHR, and TELWRKBFCVD, available since October 2022) which are unavailable among ATUS person files. To start with, I created data extract from cross-sectional CPS 2020-2024 and rectangle ATUS 2021-2023, keeping only the latest CPS record for each individual if they have multiple ones. I chose CPS 2020-2024 because my ATUS file has the “year_cps8” value ranging from 2020 to 2023.

As the previous post suggested, I merged 1:1 the ATUS file with the CPSIDP variable using the CPS file, but around 8,548 samples out of the total 25,771 ATUS samples were not matched at all. This is wired because all my ATUS records have their PRESENCE value as “both atus and cps” and should all have corresponding ATUS records. What’s worse, all samples interviewed by CPS later than October 2022 were unmatched, and all matched samples came from September 2020 to September 2022. This was terrible, since my teleworking variables were only available after Oct 2022, and all matched samples were actually useless for me.

I pasted my STATA codes here just in case. Can you tell me what’s wrong here?

**================ IPUMS CPS file 2020-2024
use "D:\Data\American Time Use Study\Linkable CPS 2020-2024\cps_00016.dta", clear
order cpsidp mish
format cpsidp %20.2g
gsort cpsidp -mish
by cpsidp, sort: gen keep_flag = (_n == 1)
forvalues i=1(1)3{
	by cpsidp: replace telwrkbfcvd = telwrkbfcvd[_n+`i'] if missing(telwrkbfcvd)
	by cpsidp: replace telwrkdiffcvd = telwrkdiffcvd[_n+`i'] if missing(telwrkdiffcvd)
}
keep if keep_flag == 1
ren telwrkpay telework
la var telework "telework 0/1 cps"
ren telwrkhr twhour
la var twhour "telework hours cps"
drop keep_flag
ren year year_cps8
ren month month_cps8
contract cpsidp mish year_cps8 month_cps8 telework twhour telwrkbfcvd telwrkdiffcvd
save "D:\PhD\Research\Teleworking Lose Community\Data\linkable_cps 2020-2024.dta", replace

**================ link files
use "D:\Data\American Time Use Study\IPUMS ATUS 2021-2023\atus_00006_person.dta", clear
format cpsidp caseid %20.2g
merge 1:1 cpsidp using "D:\PhD\Research\Teleworking Lose Community\Data\linkable_cps 2020-2024.dta", keep(match master)
tab month_cps8 year_cps8 if _merge==1
tab month_cps8 year_cps8 if _merge==3
2

Thank you for your patience in awaiting a response. I can confirm there is an error in the CPSIDP values in the 2023 ATUS sample. You should not attempt to use the CPSIDP values in the 2023 ATUS data at this time, and unfortunately will not be able to link CPS records to 2023 ATUS records using the existing data. We are working to fix the error as soon as possible. Please feel free to follow up here or by email at ipums@umn.edu if you do not see information about a fix posted on the IPUMS ATUS revision history by January.

We would like to offer you an IPUMS mug to thank you for bringing the error to our attention. Please contact ipums@umn.edu with a mailing address where we can send this small token of our appreciation.

3

Thank you, Isabel, for the update and for confirming the issue with the CPSIDP values in the 2023 ATUS sample. I appreciate your team’s quick response and the effort to fix the error. I will hold off on using the 2023 ATUS data for now and will keep an eye on the revision history for updates.

Also, thank you for the generous offer of an IPUMS mug! I’ll be sure to reach out with my mailing address seperately.

Thanks again for your support, and I look forward to hearing more once the error is resolved!

Best,

Qianyi (Sinyee)