Linking CPS Education Supplement to Voter Supplement

Hi, I have a quick question. I am doing the CPS Voter Supplement, but I also need the information on the individual current level of school enrollment. I found the CPS Education Supplement provides this information IPUMS CPS: descr: EDGRADE. So, I was thinking about merging the variable “edgrade” to Voter Supplement by using the “cpsidp” or “cpsidv”. Is it feasible?

Thanks in advance for any help that could be provided.

The merge you are describing is possible. I will provide you with some considerations necessary for performing this linkage and present you with an alternative. First, note that not all respondents to the Voter Supplement also responded to the Education Supplement. The two supplements are fielded in different months and have different universes. See this page on linking the CPS and our rotation pattern explorer tool to understand how linking respondents between CPS months works, and to determine what share of the analytical sample from the Voter Supplement would have a response to the Education Supplement. Weighting a sample that uses data from multiple supplements is a bit complicated; the most promising method is to create your own longitudinal weight that takes into account the probability of being observed in both samples. If you are interested in pursuing this method, I can provide more information about how to go about creating such a weight.

The alternative option is to use the variable SCHLCOLL, which is part of the BMS, and therefore available for all supplement respondents in all samples (as long as they are in the universe of the variable). SCHLCOLL indicates whether respondents age 16 to 24 (or 16 to 54 for ASEC 2013 onward) were enrolled in high school or college during the previous week, and, if so, whether they were enrolled full- or part-time. I am not sure whether this variable would be appropriate for your purposes. It is much less detailed than the variable EDGRADE, but its use with Voter Supplement data would be more straightforward.

Many thanks for the response! As I need information on education enrollment records for individuals whose ages are younger than 18 years old, the variables related to education in the BMS are not that ideal to me. Would it be possible for you to provide more information on creating the weight? I appreciate your help much.

Because being observed in both the Voter Supplement and the Education Supplement depends on being eligible to be surveyed in both months and not missing one of the CPS interviews (i.e., moving between the months these supplements were fielded or not responding for other reasons in October or November), you might consider adjusting the weight for one of these supplements to a longitudinal weight to account for these factors. Longitudinal weights account for differential probability of continuing to be observed in subsequent survey months. For example, you can convert EDSUPPWT to a longitudinal weight by modifying and running a version of the do file that will rake EDSUPPWT (see this page on linking and the CPS). In the Stata script, you will need to replace WTFINL with EDSUPPWT. I am not aware of anyone using, testing, or verifying this method, but this is the solution we recommend starting with when users wish to use variables from one supplement in analysis of data from another supplement.

Hi Isabel,
I see. Many thanks for your help!