CPS or ASEC weights for motherhood status by medicaid status

I am hoping to examine the share of women aged 18-45 on Medicaid that are childless in each state. I am using the person variables from the ASEC on age, sex and Medicaid coverage last year to create my category. The state of residence is an ASEC household variable. I am using the fertility supplement from the June CPS to determine whether the respondent has had a child. However, I am not sure whether I should be using ASEC person weights, ASEC household weights, or fertility supplement weights (equivalent to basic CPS weights).

Upon closer examination, it seems that there is no overlap in observations (I cannot merge based on cpsid cpsidp statecensus year pernum month serial). Any advice would be greatly appreciated!

Thank you in advance!

Camille Wixon

You should be able to link respondents of the ASEC to their Fertility Supplement responses. Respondents to a Fertility Supplement in a given year can be linked to their ASEC responses from the same year and to their ASEC responses from an adjacent year, depending on what month in sample (MISH) they are in during the month of the Fertility Supplement. See the CPS rotation pattern explorer tool to visualize the links between different CPS samples, and this page on linking and the CPS for more information on linking individuals across multiple CPS samples. Using the CPS rotation pattern explorer tool, taking the June 2022 Fertility Supplement as an example, you can see that 25% of 2022 ASEC respondents, 12.5% of 2021 ASEC respondents, and 12.5% of 2023 ASEC respondents can be linked to the 2022 Fertility Supplement. Narrowing down your sample even further by selecting on sex, health insurance coverage, state, age, and specific fertility values may result in a very small final sample.

I would recommend merging on CPSIDV, the IPUMS-validated unique person identifier. This identifier links individuals across CPS samples, and only retains links where IPUMS believes, based on the sex, age, and race values, that the link identifies the same individual over time.

There is no official guidance for weighting two CPS supplements together, but I will provide you with some guidance based on my knowledge of CPS weights and weighting guidelines. You should review the literature in your field to determine if others have used these two supplements together and how they weighted their data.

It sounds like you define who is in your analytical sample based on the Fertility Supplement. Therefore, I would recommend using the Fertility Supplement weight, FRSUPPWT. However, since being observed in both the Fertility Supplement and the ASEC depends on being eligible to be surveyed in both months (i.e. being in month-in-sample one in March) and not dropping out of the CPS (i.e., moving between the months these supplements were fielded or not responding for other reasons in February), you might consider adjusting FRSUPPWT to a longitudinal weight to account for these factors. Longitudinal weights account for differential probability of continuing to be observed in subsequent survey months. You can convert FRSUPPWT to a longitudinal weight by raking FRSUPPWT by running a version of the do file linked on this page on linking and the CPS. In the Stata script, you will need to replace WTFINL with FRSUPPWT.