Hi, I know to use the WTVET weight instead of WTFINL when using the VETSTAT variable (Veteran Status) in CPS. But how does this work with the ASEC? E.g. for veteran migration stats and other analysis only available in the ASEC, is it legitimate to use VETSTAT with ASECWT?
I am going to share some broad information about weights in the CPS in addition to answering your specific question about VETSTAT. Please read our page on weights in the CPS for more information on choosing the right weight.
Most analyses based on individual-level ASEC data should use the ASECWT variable. For analyses of the ASEC focused on household-level variables, researchers should use the household weight, ASECWTH. The variable VETSTAT is an ASEC variable, so you should use ASECWT to analyze VETSTAT.
When analyzing data from non-ASEC supplements, you should use the supplement-specific weight. For example, when analyzing data from the Veterans Supplement, you should use the weight VTSUPPWT. The variable VETSTAT is not part of the Veterans Supplement; it is part of the ASEC.
If your study population or the analytical sample for a particular analysis is veterans, you should use WTVET. WTVET ensures that estimates of parameters are representative of the U.S. population of veterans. You should not use this variable to, for example, determine the share of the U.S. population who are veterans, or for other analyses that include both veterans and non-veterans. WTVET is available in most Basic Monthly Survey (BMS) data and supplement data. It is not available in ASEC samples. In Veterans Supplement data, WTVET is equal to VTSUPPWT.
Most analyses of Basic Monthly Survey (BMS) data should use WTFINL for person-level analyses, and HWTFINL for household-level analyses.
Hi Isabel, thanks for your response. Can you clarify your note that VETSTAT is an ASEC variable - isn’t it both a CPS and an ASEC variable as seen here?
Yes, VETSTAT is available in BMS data as well. From your original post, it sounded like you were analyzing ASEC data. The survey question we use to create VETSTAT is asked in the BMS. IPUMS CPS extracts of ASEC data also include variables from the March BMS record. When you analyze ASEC samples, you are analyzing a sample population that differs from the BMS sample population, and therefore need to use sampling weights that are specific to the ASEC sample.
- If you are analyzing VETSTAT in BMS samples, you should use WTFINL for person-level analyses, and HWTFINL for household-level analyses.
- If you are analyzing VETSTAT in ASEC samples, you should use ASECWT for person-level analyses, and ASECWTH for household-level analyses.
- If your study population or the analytical sample for a particular analysis is veterans, you should use WTVET.
Thanks Isabel. Can you offer any guidance on the combined used of VETSTAT together with another person variable like AGE (e.g. count of veterans in a certain age range) - is it legitimate to use WTVET with both?
The variables VETSTAT and AGE are both available in ASEC and BMS samples, so I cannot definitively say which weight you should be using. The weight to use with your analysis depends on the samples you’ve chosen to include in your extract and analyze. You can determine which samples you are using by examining the variables YEAR, MONTH, and ASECFLAG. Or you can view the samples included in each of your IPUMS CPS extracts under “my data.”
Here is a summary of which weight to use depending on the samples you are analyzing and if veterans are your study population:
- If you are using data from the ASEC, you should use the ASEC weights ASECWT for person-level analyses, and ASECWTH for household-level analyses.
- If you are using data from the BMS, you should use the basic monthly weights WTFINL for person-level analyses, and HWTFINL for household-level analyses.
- If you are using data from a non-ASEC supplement, you should use the weight specific to that supplement.
- If your study population (the group of people included in your analysis) is veterans, you should use WTVET. Only use WTVET when you are not including non-veterans in your analysis. This weight is for studying the veteran population exclusively.