Trying to match ftotval and hhincome to published income data from the Census

I am sure I am making an elementary mistake, but when I calculate the median and mean of ftotval, it is a lot lower than the mean and median family income from the Census and vice versa the mean and median for hhincome are much higher than what the Census tables show. I am comparing nominal values to the published tables for 2021, so deflation is not a factor (I think). I am also using asecwt. Any pointers on what I may be missing? Thanks.

I figured out my mistake. Sorry to bother you.


Hello @Christian_Weller, are you able to please explain the mistake you made and how you resolved this problem? I am a very novice user of IPUMS CPS data and am having the same issue. I pulled HHINCOME, took a simple median, and for year 2022 (survey year 2023) the value is $91,200, whereas in the Census report “Income in the United States: 2022”, nominal median household income for 2022 was $74,580. I’m sure my mistake is also extremely elementary…

I used all observations instead of just one per household. That was my biggest error if I remember correctly. I think I also used the wrong weight. I think it should be asecwth.

1 Like

Hi Evan. Like Christian stated, when performing analyses of household level variables, such as HHINCOME, you should exclude your analysis to just one person per household. You can do this easily by filtering PERNUM=1. You also need to use a household weight for household level analyses. The household weight for the ASEC is ASECWTH. We also offer replicate weights to use with ASEC data, which are necessary for estimating empirically derived standard errors or confidence intervals.

Thank you very much! That is very helpful, and this resolves the issue.