ACS Data Users Group

For an analysis of trends in the household population--specifically grandparents as caregivers for their grandchildren--are there cautions to be wary of when processing the ACS PUMS from 2000 through 2013? The obvious difference is one of sample size during the development period 2000 - 2004 and its effect on statistical reliability. Others to be concerned about?

Best,
Warren A. Brown
Cornell University
wab4@cornell.edu

Hi Warren -
Data from 2000-2004 were technically from a pilot of the ACS. Full implementation did not really begin until 2005.

Also, one of the other main changes in the survey was the inclusion of group quarters into the sample, starting in 2006. This, theoretically, should not affect household populations (e.g. grandparents caring for grandchildren are in households, not GQ) BUT it may affect the population controlling that occurs for small area estimates.

There may be other issues, but those are the two that come to mind.

My recollection is that the sample design for the national sample was quite different during the 2000-2004 demonstration period: it was a multi-stage sample with self-representing and non-self-representing PSUs. If that is the case, there would be some states with no PUMS data. Of course, if you're only interested in U.S. estimates, this probably has no effect. Also, I don't think there is any PUMA geography for the 2000-2004 PUMS. Sometime around 2000, but I don't remember the exact year, the question about length of time a grandparent is responsible for a grandchild changed the answer categories in a manner that made it impossible to compare that question back to previous periods. If the change happened in 2000, then there is no problem going forward.

Doug Hillmer

I want to correct something I said in my reply to Warren. There WILL be PUMS data for every state, but the sample design during the demonstration period allows only national and state-level estimates. Sorry about that mistake.

Doug Hillmer