I calculated the higher education enrollment rates at state level using ACS from 2001 to 2012. I know that since 2006, the ACS captures population living in group quarters. Given the topic that I am analysing (enrollment in higher education of multiple population groups), how can I make appropiate comparison between ACS pre and post 2006? Should I work only with information from individuals living in households, i.e., those with gq equal 1 and 2?
Also, how appropiate is to make comparisons between ACS 2001-2004 and ACS from 2005 and beyond given the substantial differences in sample size?
It is a good idea to be cautious when comparing the 2000-2004 ACS samples to the 2005-onward ACS samples. As you noted, one of the most potentially disruptive differences between the two sets of samples is the exclusion of Group Quarters from the 2000-2004 ACS samples. It may be a good idea to remove Group Quarters from the later samples, as you suggest, to increase the consistency of the sample universe. I would recommend trying a few analyses with and without the Group Quarters exclusion to try to gauge the impact. The 2000-2004 ACS samples are also smaller than the 2005-onward samples (as noted on the Samples page), which makes them less likely to capture smaller sub-populations and specific geographic areas. These differences may impact your analyses, but ultimately it is up to you to decide if these differences are acceptable for your study.