Question for 'RANK PROXIMITY SWAPPING' & SPM

Dear team. I have two questions related to the poverty measure variables.

Firstly, could you provide more information or recommend articles on the rank proximity swapping method? I imputed some income variables of other panel data(Health and Retirement Study) and tried to do a sensitivity check using CPS-data. Therefore, I need to do top-coding similar to the CPS method. However, I failed to find a more specific way to do rank proximity swapping. (I already read Larrimore et al.(2008) and some instruction(https://cps.ipums.org/cps/topcodes_tables.shtml), however, the focus was a little bit different.) To be specific, I read that the swapping occurs within a ‘bounded interval’ so I am wondering what the bounded interval means.

Secondly, I am wondering whether SPM medical out of pocket expense variable(name : SPMMEDXPNS) include the amount of health insurance premiums(variable name : HIPVAL) or not. I heard that SPM medical out of pocket expense ‘conceptually’ includes health insurance premiums, however, some sample’s SPM medical out of pocket expense doesn’t seem to include health insurance premiums when I analyzed CPS data.

Thanks a lot!!

According to this Census Bureau report (p. 3):

The bounded interval is large enough to include many people/households in order to protect the data and small enough to ensure that the swapped values are within proximity of each other. The parametric details of this are confidential.

You could choose your own bounded intervals above the published threshold values based on those criteria and the result probably won’t be terribly different.

Regarding SPMMEDXPNS, this includes premiums, as does the related variable MOOP. I verified this with the original microdata from the Census Bureau and NBER, which includes some variables that are not currently available on IPUMS CPS. These files contain variables that break down OOP expenditures by premium, medical care, and OTC (MOOP, PHIP_VAL, PMED_VAL, POTC_VAL). You can find these variables on the datasets available at NBER. If you’d like to append these variables to your IPUMS extract, you can do so by following the instructions in this blog post. You might have seen apparent discrepancies because SPMMEDXNPS uses a different family definition than HIPVAL. However MOOP uses the same family definition as HIPVAL.