Records that are not in IPUMS ATUS dataset

I’m using the weight variable in the ATUS-CPS data from BLS website and merging it into the IPUMS ATUS data 2004-2018. I find 7 records that are in BLS dataset but not in IPUMS ATUS dataset. Here are the CASEID of these records.


Since I see the base weight for the case “20060706061701” is -1 (but not the rest of the 6 records), I wonder if IPUMS purposefully excluded these records. If so, could I find the reason of the exclusion somewhere on the website?


Thanks for this note. I am, unfortunately, not able to replicate your observation. Although I also do not find any observations with these CASEID values in the IPUMS ATUS data, I also do not find any observations with these TUCASEID values in the single or multi-year ATUS files on the BLS website. Is it possible you are using data from a different BLS page, file, or source? Otherwise, I’m not sure what is happening that is causing these values to show up in your data.

Thanks for the debugging.
Sorry I didn’t explain it explicitly. I’m using data including both respondents and nonrespondents from IPUMS ATUS and the “ATUS-CPS 2003-2018 file” from the BLS website (

I think these 7 cases are all nonrespondents. Did you include nonrespondents in your searching?

Thanks for this additional information. I still do not have a good answer to this question. Could you send me a link to the page where you found the data file you link to above?

Hi Jeff,

Sorry to hear that and thanks for the efforts you made.

Here is the link.

It is the “ATUS-CPS 2003-2018 file(zip)” that I used. I took a screenshot of part of the page and circled the file I clicked and downloaded.

After unzip the file, I used the SAS version of the datafile.

I tried to search my original datafile downloaded from this website again. I tried 2 TUCASEID of these 7 records and they are included in the datafile.

Let me know if you have any questions.



Okay, I see where the file is coming from now. Thanks for the link! I’ve summarized this observation and passed this along to the IPUMS Time Use Team to investigate. I suspect this is due to a very obscure detail in the way the data are harmonized on the IPUMS platform. I’ll keep you updated with any developments.

Sounds good. Thanks Jeff.

The reasoning for these missing observations is as follows: The one case from 2006 was dropped because it had incomplete information across files. The six cases from 2010 were dropped because they had the identical CPS identifiers to other records in the ATUS-CPS file. Each of these observations were non-respondents and had identical TUBWGT values to the records that were retained. I hope this helps.

It makes sense to me. Thanks for the clarification, Jeff!