Why are there apparently too few voters in the 2002 CPS Ipums voting and registration file?


I am analyzing the 1996-2008 CPS Voting and Registration supplements, and have been able to replicate a number of Census reports but had trouble with 2002. I used the iPums data and syntax generating system –

The iPums file shows 87,762,207 reporting voted in the 2002 election (using WTFINL)

Whereas the Census tables show 88,903,000 (last 3 digits not reported) - that table is Table 2-1 on this page http://www.census.gov/hhes/www/socdem…

My calculation of the citizen voting-age population appears to be the same in both iPums and the tables, so the discrepancy appears limited to those reporting to vote.

Any insight in to this apparent discrepancy would be appreciated


When CPS files are initially released and summary tables are generated they are weighted based on information from the most recent Decennial Census. However, it can take 3 or 4 years to generate new estimates and weights from a Decennial Census, so files near the beginning of a decade will be revised once the new weights are created. Sometimes the summary tables are not updated making it impossible to match the published tables. I believe this is the case for the 2002 Voter supplement. This is evidenced by the smaller Total 18+ population in the weighted IPUMS-CPS sample (roughly 207 million) compared to the Total 18+ population in the published table you linked to (210 million). The IPUMS-CPS 2002 sample uses the more current weights (revised based on the 2000 Census) and the table was likely created before the new weights were applied to the data.

I hope this helps.

Thanks very much for this response. Is there documentation on this site that confirms the public use file weights are based on the 2000 Census? I was a little surprised the CPS published tables might not use such weights since they were released on the Internet in 2004 (though they may have not been revised since their initial release).


There is a discussion of the updated weights in the Source and Accuracy report for the 2002 Voter supplement. The updated weights were released in January of 2003 (according to NBER). And, though I was correct about the source of the difference between the published table you linked to and the data in IPUMS-CPS being 2000 population adjustments, it appears that the IPUMS-CPS data is the party using the old weights. The data set is being updated to include the new 2000-based weights presently.

In the mean time, since you pointed out a significant error in the IPUMS data you are now entitled to a shiny new IPUMS MUG. Please send an email to ipums@umn.edu with a mailing address where you can receive your coveted prize. Thank you for bringing this error to our attention and helping us maintain this important data.