How to determine top coding / cutoff for HHINCOME

I’d like to use the IPUMS-CPS microdata to generate national estimates of household income at the median, at the 90th percentile, 95th percentile, and so on, and I’d like to figure out the highest percentile I can reliably calculate with this dataset (98th? 99.99th?, etc). I’m using the HHINCOME variable (and collapsing individuals into households, weighting by hwtsupp, adjusting for inflation), and I’m trying to produce annual estimates from 1980 to 2012. I’m having trouble finding the right documentation on how HHINCOME has been top-coded.

I see in my dataset that as of the 2013 (March AESC) data, the max HHINCOME value is 2,742,997 in current dollars. I don’t seem to be picking up any 999999’s, N/A’s or other figures that clearly look like they’ve been replaced.

I’d appreciate it if anybody could point me in the right direction!

HHINCOME itself is not top-coded, but it is created using personal income (INCTOT), which is itself created from top-coded income components. The ‘Codes’ page for INCTOT contains a link to a User Note that explains the top-coding process, as well as a link to the actual Topcode Tables for each year. As you will notice, each component of INCTOT is top-coded separately.

Keep in mind, a value of ‘99999999’ for HHINCOME is the NIU coding, but there is no top-coding specific to the HHINCOME variable.

I hope this helps!