I am trying to construct gross flows by occupation (i.e., gross flows for each occupation group for individuals linked across two consecutive months in the basic monthly CPS) and I would be grateful for any guidance on which weights I should use for this application.
In particular, while I have used PANLWT to successfully replicate gross flows for the entire (civilian) population, my question that I would like to discuss is whether PANLWT is appropriate for constructing gross flows for subpopulations (e.g., gross flows for occupational groups) or if these weights are meant only for aggregating survey data to construct gross flows estimates for the entire population. If not, do you have any insights into how I should go about constructing the ideal weights for this application?
As discussed in the previous thread, PANLWT is appropriate for gross flows analyses in general. I would encourage you to read pages 85-86 of Technical Paper 66, which discusses longitudinal weights in the CPS. It specifically states:
These longitudinal weights reflect the technique that had been used prior to 1994 to inflate the gross flow esti-mates to appropriate population levels. That technique inflates all estimates or final weights by the ratio of thecurrent month’s population controls to the sum of thesecond-stage weights for the current month in the matched cases by sex.
There is no mention of adjustment by occupation or any other characteristic. That being said, PANLWT is derived from WTFINL, which is adjusted for demographic characteristics. My best suggestion is contact the Bureau of Labor Statistics directly to see if they have any guidance on this issue; I would also recommend reviewing the literature in your field to identify how other researchers have handled this. Finally, if you are interested in creating your own longitudinal weights, you may be interested in the weighting resources for linking IPUMS CPS data.
For anyone interested, here is the response I received from the BLS:
general advice on estimating gross flows:
The longitudinal weights provided on the CPS public use files do not [exactly] replicate the labor force flow series produced by the BLS. The BLS research series on gross flows [published here] includes estimates of population inflows and outflows, such as those due to migration and deaths, which cannot be estimated directly using just CPS data, as well as other sources of discrepancies between the flows and the monthly stocks. (See “Estimating Gross Flows Consistent with Stocks in the Current Population Survey” and the discussion of PWLGWGT in the Monthly CPS Public Use Microdata Files documentation.)
direct answer to my question here on estimating gross flows by occupational groups:
At any rate, the longitudinal weights provided may be used for flow estimates for occupational groups recognizing that they ignore movement into and out of the civilian noninstitutional population and rotation group effects, and that for small groups the estimates will be quite noisy.
