It looks like FOODSTMP and HEATSUB should both have the same universe of households (i.e. all interviewed households and group quarters). However, there seem to be 177 individuals who were listed as NIU for HEATSUB but who had a valid response for FOODSTMP. These individuals occur in the ASEC file between 1982 and 1987 and the QHEATSUB flag is either set to missing or not allocated. In the case-count view of the documentation, these variables also report a different number of households as NIU for these years. Was there some difference in which households were asked about heating assistance in those years?
It is often the case that the universe asserted by the documentation is not fully enforced in the original data for pre-1989 samples. The informal policy of IPUMS CPS is not to enforce universes where only a few cases per sample are outside of the universe documented by the BLS. Which is why these 177 records across 5 samples maintain their meaningful values even though they fall outside of the question universe.
Our general suggestion is to either recode the “no” values in FOODSTMP to NIU for those records that are NIU in HEATSUB. Or simply drop those records from the analysis. Either way, these cases are a relatively small share of each of the five samples.