How to use CPS Food Security weights for summary statistics

CPS
1

Hi. I just got in touch with the CPS data and I want to calculate some summary statistics, for example, the % of households falling into different security categories for each race.

I realized I should use weights when doing such calculations, but I’m confused about how to do so.

Here is my approach for calculating the statistics for a single year but the numbers I got seems off:

  1. subset the data for 2019

  2. keep the records that finished the survey by dropping the records with FSSUPINT==2

  3. keep only one record per household by dropping the duplicates based on CPSID

  4. calculate the weighted food security scales by multiplying HWTFINL and FSSTATUS

  5. sum up the weighted FSSTATUS of records whose RACE==200 (Black) and FSSTATUS is not equal to 98 and 99. This is the sum for all black households in the universe and did respond

  6. sum up the weighted FSSTATUS of records whose RACE==200 and whose FSSTATUS==2 (Low food security)

  7. divide 6 with 5. This gives me 0.18, which is off what’s shown in USDA ERS - Interactive Charts and Highlights

Could anyone let me know how to properly use the weights when doing the specified tasks? Thanks in advance.

2

I’ll give some comments on your steps below:

  1. subset the data for 2019
  1. keep the records that finished the survey by dropping the records with FSSUPINT==2
  1. keep only one record per household by dropping the duplicates based on CPSID

Since you are looking at RACE, but with a household level analysis, it is common to use the race of the householder as a proxy for the “race” of the household. I’m not sure if that’s what USDA used, but it probably is. To do this, keep only those with PERNUM=1 (instead of dropping duplicate CPSID).

  1. calculate the weighted food security scales by multiplying HWTFINL and FSSTATUS

You shouldn’t do this if you’re treating FSSTATUS as a categorical variable, which it sounds like you are.

  1. sum up the weighted FSSTATUS of records whose RACE==200 (Black) and FSSTATUS is not equal to 98 and 99. This is the sum for all black households in the universe and did respond

You should be summing HWTFINL instead of FSSTATUS

  1. sum up the weighted FSSTATUS of records whose RACE==200 and whose FSSTATUS==2 (Low food security)

Again you should sum HWTFINL instead of FSSTATUS

  1. divide 6 with 5. This gives me 0.18, which is off what’s shown in USDA ERS - Interactive Charts and Highlights