I am attempting to replicate the CPS file POV46 - Sample - All Ages by state using the IPUMS data for March 2013 to get the same numbers of poor households per state. I assign poverty status on my own based on the 2013 poverty thresholds and ages of household members but I fall short of the CPS numbers. If I instead use the offpov variable the numbers still fall short. I also see that there are 68,000 addresses in the 2013 CPS data (with another 30000 asked redesigned income questions for 2014) and it looks like about 74000 in IPUMS-CPS. Does IPUMS use a different sample of survey respondants than CPS?
I am afraid I am not able to replicate your results. The OFFPOV variable should replicate the official poverty rate statistics but the POVERTY variable is not expected to, as discussed in the Replicating Official Poverty Rates user note. When I compare a cross tabulation of the IPUMS-CPS 2014 March ASEC STATEFIP by OFFPOV (unweighted counts) I am able to match the All Ages table exactly (likewise the IPUMS-CPS 2014 March ASEC frequencies weighted by WTSUPP also match the Census Bureau’s weighted frequencies as well). I am also not sure where you are seeing the 74,000 addresses in the 2014 IPUMS-CPS ASEC file. The 2013 ASEC included 74,821 inhabited households. Is it possible that you are using the 2013 data?
I hope this helps.