Data on retirees returning to the workforce

I am pretty sure that there is not but sometimes one gets surprised. Is there data on this set that tells you how many retirees have returned to the workforce for a given time or place (state as an example).

This is definitely a question that can be answered with the CPS, but it will require some manipulation of the data. Specifically, it will require leveraging the panel component of the CPS to observe individuals over time. I will provide a bit of information on the panel aspect of CPS, how to think about transitions out of retirement, variables to use to get timing and geographic information, and some comments about making population estimates.

Individuals may be included in the CPS up to 8 times over a 16-month period. IPUMS CPS provides pre-linked data files linking individuals who participate in the annual socioeconomic supplement (ASEC) between their first and second ASEC interview (this file will capture the person at two time points); however you can manually link individuals across months using CPSIDP (this approach will yield up to 8 observations). Once you have linked individuals, you will look for instances where the person’s EMPSTAT value changes from “NILF, retired” (EMPSTAT = 36).

If you’re only interested in transitions from retirement to having a job, you will want to note cases where EMPSTAT changes to “At work” (EMPSTAT = 10) or “Has job, not at work last week” (EMPSTAT = 12). If you’re interested in transitions from retirement into the labor force regardless of employment status, you will also want to include cases where EMPSTAT changes to “Unemployed, experienced worker” (EMPSTAT = 21) or “Unemployed, new worker” (EMPSTAT = 22). You can use the MONTH and YEAR variables to identify when this transition occurred. You can then use a geographic identifier such as STATEFIP to estimate the total number of these transitions for a given place.

Since people transitioning out of retirement in a specific time period or in a specific location is likely to be a small population depending on how granular your place and time specifications are, I recommend keeping a close eye on the margin of error for your estimates. Note that using CPS data longitudinally requires different weights to get correct population estimates. If you are linking respondents across the full 8-month panel, you will want to use the longitudinal weight LNKFW8WT. You can find other longitudinal weights here for alternative research designs such as linking respondents across adjacent months or between two years.