Question

Status:Closed    Asked:Jun 22, 2016 - 10:00 AM

What weights do I use to analyze combined IPUMS CPS and Census (1960-70) data for occupation?

 
Do you have the same question? Follow this Question
 

Staff Answer

avatar

Joe_Grover

Staff

Since both the IPUMS-CPS and IPUMS-USA samples are weighted to represent the entire national population pooling the IPUMS-CPS 1960 ASEC sample and the 1960 Decennial Census from IPUMS-USA will result in a national population that is two-times too big. To account for this the easiest adjustment is to divide each person's weight by 2. There are also multiple methods available for re-weighting the data if you are interested in a specific sub-population or want to preference one data source over the other. When pooling data from two different data sources, we recommend that users be aware of the different collection/weighting methods used by each data source. More information on sample designs are available in the IPUMS-CPS and IPUMS-USA documentation. Another way to utilize the compatibility between IPUMS-USA and IPUMS-CPS is to use annual estimates from IPUMS-CPS to fill in years between the decennial censuses from IPUMS-USA. Variables like OCC1950 are coded to be comparable between these two data sources, but it is still important to make sure that the universes and question wording is similar enough to be comparable. For this kind of combination of IPUMS-CPS and IPUMS-USA data (i.e. filling in years between decennial censuses) you simply use PERWT for the IPUMS-USA samples to generate national-level estimates and (assuming you are using data from a CPS supplement) WTSUPP for the IPUMS-CPS samples.
I hope this helps.

 

Jun 22, 2016 - 03:51 PM

0
0
Report it

OTHER QUESTIONS NEEDING ANSWERS

Can I analyze IPUMS-CPS data without a statistical package?
pooling ipums USA and ipums CPS
Login   |   Register


Share |