I have read through the existing responses and several articles re: merging CPS March and monthly supplements over the years, yet am still left struggling. Essentially I am trying to look at how job tenure impacts income/employment in older/less healthy workers from 1989-2013. The problem that I’m finding is that I am using the March Supplement 1989-2013 for its income/employment/health information as my main dataset, but the job tenure data is only located in the Jan/Feb supplement of years 1995, 1997, 1999, 2001, 2003, 2005, 2007, and 2009. The job tenure information is more or less useless to my project unless I can merge it with corresponding income/employment/health information. How would you recommend going about this? Do you think it would be worthwhile (or even possible) to try and track individual respondents over time and merge that way?
Thank you in advance for any help you are able to provide.
Based on the rotating design of the CPS survey, individuals are interviewed in the same four-month period of two consecutive years. Unfortunately, since the Job Tenure supplement occured every other year, respondents will have the opportunity to answer the Job Tenure questions only once. This makes tracking respondents over time impossible for this set of questions.
However, it is still possible to link the Job Tenure and March supplements within a given year. Linking CPS monthly samples prior to 1994 is difficult; however, the Job Tenure variables only go back to 1995. You can find the identifiers necessary to link the non-March supplement data to the March supplement data here. The March supplement has different identifiers than the non-March supplements. In order to link at the person level, you will need identifiers not available in the IPUMS data. IPUMS and NBER files share a common sort order, so you can perform a sequential merge in your stats package to merge the NBER identifiers onto the IPUMS files, by year. That is, match the first NBER record to the first IPUMS record, the second to the second, etc for each sample. This will effectively add the NBER identifiers onto the IPUMS records. You can then use the NBER identifiers as described in the linked resource above.
You can also find NBER resources here that provide STATA .do files for matching across March CPS surveys, but these programs can be adjusted to perform the match you desire.
Hope this helps.
Thank you for your thorough response- It got me on the right track. I am having some trouble running the sequential merge for the 1995 March Supplement. For some reason the NBER portion is telling me there are 285,550 responses for March 1995, and the IPUMS version is telling me there are 149,642 for March 1995. To confirm, in order for the sequential merge to be successful, there should be an equal number of responses, correct? In addition, is it necessary to type a “sort” command prior to running the “merge 1:1 _n” command?
Thank you, again!
One clariyfing point to add to the above message-- Whenever I try to add on the NBER identifiers to my existing IPUMS CPS datatset, there are always a substantially higher number of NBER observations than IPUMS. This is the case for multiple years. For 1997, there are 131,854 IPUMS observations, but 252,663 NBER observations.
Thank you, again, and sorry for the confusion.
I apologize for the back and forth. Since my last response I have figured out the error on my end and am all set for the time being. Thank you again for your help!