I just wanted to make sure that I understand the basic CPS and outgoing rotation group sample design correctly.
Every individual is interviewed for four months, then is not interviewed for eight months, and then interviewed for another four months. In every month the individual is asked the basic questions and depending on the calendar month the individual is also asked specific supplement questions.
As for the outgoing rotation group, any individual in the 4th or 8th month belongs to the outgoing rotation group and not only individuals that belong to ASEC? So individuals in the outgoing rotation group have been interviewed in any calendar month and not just the March? Therefore the sample size of the outgoing rotation group is bigger than of ASEC?
You are correct that all individuals in the 4th and 8th rotation groups are in the ORG/earner study. This is true for all months, and is independent of ASEC (which is collected only in March, with some caveats). You can see the rotation group of a given household using the variable MISH.
Regarding sample size, the sample size of the ORG in any given month is approximately 1/4 of the sample in that month. If you add up the ORG responses over the whole year, there will be more cases than in the ASEC for that year. But the size of the ORG sample in March will be only about 1/4 as large as the ASEC sample. Note that most questions in the earner study refer to the previous week only, and so are not closely comparable to questions in ASEC, which mostly refer to the entire previous calendar year.
Got it! Wow, the universe of ORG - civilians age 15 and older who are currently employed as a wage or salaried worker (that is, not self-employed) is decreasing the sample by so much to 1/4? Or are there any other sample restrictions for ORG?
In any given sample, only 1/4 of households are in rotation groups 4 or 8. The remaining households are in rotation groups 1-3 or 5-7. On top of that, there are universe restrictions for the ORG questions, as you mentioned. These reduce the sample of individuals to less than 1/4. This link gives more information on the CPS rotation pattern.
Hi, I had a question that piggy-backs off of the ones posed in this thread. Are the individuals in the ORG also eligible or part of the sampling for the ASEC? For example, are there individuals asked questions from ORG and ASEC? In downloading the dataset, this seems to be the case, but just want to be sure, as I couldn’t find a direct answer to this online. Additionally, when using a blend of variables (for example in a survey weighted generalized linear model) from both ORG and ASEC, is it appropriate to use survey weight EARNWT only?
I appreciate any guidance you may be able to offer. Thank you kindly.
Many of the ASEC respondents are also asked the ORG questions. However the values of MISH in the ASEC are not always accurate for oversampled individuals, so you cannot rely on MISH to identify these respondents. Instead use ELIGORG. Note that UHRSWORKORG, WKSWORKORG, and OTPAY are not included in the ASEC data after 1994 (though for people who match to the March Basic sample, these variables can be merged in from the March Basic sample, using CPSIDP as a linking key).
When conducting analyses using the ORG variables, EARNWT is generally the appropriate weight to use. You don’t need to modify the weight for ASEC samples, because the ORG respondents are a proper subset of the ASEC respondents.