IPUMS CPS ORG versus NBER CPS ORG for 2018

I find that the sample size for 2018 IPUMS CPS Outgoing rotation group (ORG) data is about 10 percent of the 2018 NBER CPS Outgoing rotation group data. Is there an explanation to why the sample sizes are different?

If, however, the mismatch is because of how the IPUMS CPS ORG data is downloaded, I have used the following criteria – Downloaded 2018 ASEC sample with ORG variables, and used the following conditions to isolate the ORG sample – Age = 15+; EMPSTAT = 10-12; CLASSWKR= 20-28; MISH= 4 or 8. Does this work?

Thanks for your help!

The merged outgoing rotation group file available via NBER merges all of the individual ORG basic monthly samples from a given year. So, if you want to compare IPUMS ORG data to NBER ORG data you’ll need to use each 2018 basic monthly file from IPUMS, rather than just the 2018 ASEC file.

I compared the sample sizes in the IPUMS CPS ORG files and the NBER CPS ORG files.

I found that if you take the IPUMS basic monthly CPS files and keep all people in the sample with mish == 4 or mish == 8 and 16 < age <= 64 you get the sample number of observations as the NBER CPS MORG files where 16 < age <= 64.

I tested pooled samples 2011-2018 and got 2,010,172 observation in both cases.

Restricting the IPUMS sample to cases where ELIGORG == 1, the sample size for the IPUMS sample is smaller than the NBER CPS sample.I think this is because ELIGORG == 1 imposes additional restrictions, but I am not sure.

Of the 2,010,172 observations, 759,215 were listed as “not eligible” according to ELIGORG.

Is this correct?

Yes, this sounds correct to me. As you are finding here, comparing the Outgoing Rotation Group (ORG) data between IPUMS and NBER can be a bit tricky. This is because in IPUMS we provide the ORG data as a subset of the basic monthly samples. NBER, on the other hand, provides the merged ORG data in their own dedicated files and how these data go from the basic monthly files to the merged ORG files is not well-documented. Therefore, you are correct that the ELIGORG variable currently uses a more restrictive universe than is used by NBER, but I am not able to detail exactly what these restrictions are at the moment.