Which 1970 sample should I use when comparing occupational breakdowns to recent estimates?


I am currently using the online data anlysis tool to try and compare the share of employed workers in the United States in 1970 who were lawyers to the share that were lawyers in 2012. I have two small questions - which I realize may be quite obvious - but which I would greatly appreciate your opinion on.

  1. Which 1970 sample should I be using when looking at these occupational breakdowns (1970 1% Form 1 vs. 1970 1% Form 2). The occ1990 variable is available for both, but they provide very slightly different occupational figures. Is there any reason to use one over the other, or should I be attempting to combine them?

  2. Would you discourage users from comparing the occupational percentages taken from decennial census data to those taken from ACS data, or is it considered acceptable to do so? I realize the survey methodologies are of course extremely different and the universe of workers being questioned about their occupation has shifted slightly in terms of age, but given that the occ1990 variable was made available in the recent ACS samples as well it appears implied that such comparisons can be done. Do you have any opinion on this?

Thank you so much for any help you can provide. It is greatly appreciated!



The difference between the 1970 1% Form 1 and Form 2 samples (which are the metro samples on the online data analysis tool) are the variables available in each as the samples represent two different questionnaires distributed to different households. So, it makes sense that the occupation frequencies from the Form 1 sample would be different than those from Form 2 sample. If both the Form 1 and Form 2 samples contain all of the variables you are interested in, it is your choice as to which one you use or you could use both.

As long as you account for universe differences, sample differences, and other comparability issues, we encourage the comparison of integrated variables over time. One of the main benefits of the IPUMS-USA data is the way the data facilitates comparison over large spans of time and the OCC1990 variable is an excellent example of this.

I hope this helps.