August and December 2014 CPD show zeroes for CPSID and CPSIDP variables. Why would that be?

Perhaps I have done something wrong, but August and December 2014 CPD show zeroes for CPSID and CPSIDP variables. For other months in 2014, for example April 2014, CPDID and CPSIDP appear just as one would expect. Any help would be greatly appreciated.

The zero values for CPSID/CPSIDP in August and December 2014 are a known error in the IPUMS-CPS data. We plan to release corrected data within the next month.

Hope this helps.

Hi, I recently downloaded the CPS ASEC and found there are still zeroes for CPSID/CPSIDP for various observations from 2008-2018. I plan to just dump those observations, but it just sucks when it’s dropping 16% of the observations I have. Just wanted to bring this to someone’s attention :slight_smile:

The CPS ASEC includes an oversample of Hispanic and SCHIPLY respondents to increase the reliability of estimates for these groups (read more in the sample design documentation). As stated in the description tab for CPSIDP, respondents that are part of the oversample are assigned a CPSIDP of 0. Additionally, the variable ASECOVERP identifies respondents that are part of the ASEC oversample. I looked at the 2008-2018 ASEC sample and verified that all cases of CPSIDP = 0 are part of the oversample. For further reading on the CPS ASEC oversample and why it cannot be linked, see page 8 of this paper, which outlines how to use ASEC data as part of analyzing CPS data as a panel.

Thank-you. What I don’t understand is that if the oversampled come from adjacent months to March why didn’t their CPSIDP transfer as well?

CPSIDP is an identifier variable that was created by the IPUMS CPS team; they did not create it for respondents in the ASEC oversample due to the fact that it is very labor intensive in some cases and impossible in others (see page 8 of this paper on Using the ASEC as part of a CPS panel for more information).