Issues Matching 2001 ASEC/March CPS to NBER data and data directly from Census


Hi all,
I’m matching the IPUMS CPS March Supplement data with TRIM3 data from the Urban Institute, which uses data directly from Census. Every year matches perfectly with the TRIM data, except 2001. I also replicated this mismatch with the NBER data. I suspect the issue is related to the addition of the SCHIP supplement. But, what I don’t understand is how the sample from Census/NBER aligns with that of IPUMS. It seems that there are some individuals in one file and not the other (so neither file is a subset of the other). Can someone clarify the difference between the IPUMS files and the NBER files? I would have expected the IPUMS to be contain all the cases of the NBER files as well as additional individuals from the SCHIP, but that does not seem to be the case



Sorry for the delay in answering this question. This took us a little bit to figure out. You are on the right track in thinking that the issue is related to the SCHIP expansion. Basically, the trick here is that there are two ASEC files available via both the CPS FTP site and the NBER site. One file includes the SCHIP cases and the other does not. As noted on this page, IPUMS provides the 2001 ASEC data with the SCHIP cases included. So, if you download the SCHIP file from NBER, you should be able to merge using HSEQ and LINENO from the IPUMS data and PH_SEQ and PULINENO from the NBER ASEC file with the SCHIP cases included.


Hi thank you this is helpful! I’m still having some issues though. Does the addition of the SCHIP cases into the IPUMS files change the hseq and serial variables for the original 2001 sample? I’m trying to merge the IPUMS file with a file from TRIM3, which is based on only the original sample. I would expect that some cases from the IPUMS file would go un-matched because of this larger sample, but there are also cases from the TRIM3/original cases that are failing to be matched with cases from the IPUMS files. Since it sounds like the original files are a subset of the IPUMS, I wouldn’t expect to have unmatched data in that direction.


After looking into this for some time, I am not sure there is any way to merge the non-SCHIP NBER ASEC 2001 sample onto the IPUMS ASEC 2001 (which includes the SCHIP). I also find that there are both non-matched cases included in the IPUMS data and non-matched cases included in the non-SCHIP NBER data. This is when using HSEQ and LINENO as linking keys. Additionally, AGE and SEX do not match for the matched cases.

The HSEQ variable is just the household sequence within a given file, therefore HSEQ values are not comparable between the SCHIP and non-SCHIP samples. Therefore, unfortunately, I do not think it is possible to merge the non-SCHIP NBER ASEC 2001 file onto the IPUMS ASEC 2001 file.


Hi Jeff,
Thank you so much for looking into this so in-depth! I really appreciate the time and it’s good to know that it wasn’t just my mistake.