Diffferent sample sizes across time

#1

Hi. I am trying to recreate and extend the results from a study published in 2002 that used, according to the study, 1970, 1980 and 1990 1% PUMS data from the “United States Census”. One of the important variables is METAREA. Initially, I thought they did this in order to only use the decennial census. In order to extend the study, I’d like to add 2000 and 2010 data.

Since I’m a total noob, are they referring to ACS, AHS, the Metro sample or some other sample?

Also, it seems like one cannot get this variable from only 1%-samples? What implications does this have? Using the exact same data would be nice, but I think as long as I get coefficients of similar magnitude and significance, that’d be fine.

A bit lost at this early stage. Best regards

0 Likes

#2

METAREA is available for the 1970 metro sample and the 1980 and 1990 1% Decennial samples. The 1% samples do not appear by default, so to have them appear on a codes page they need to be selected manually from the sample selection page.

As you’re replicating the study, the importance is not so much that you use samples of the same size but instead that you apply the proper weights. Applying weights will make a sample (regardless of whether it is a 10% sample or a 1% sample) representative of the U.S. population. If you wish to continue using 1% samples, those are available for the 2000 Decennial and the 2010 ACS.

I hope this helps!

0 Likes