I am looking at the distribution of negative values in the ACS variable ftotinc between 2001 and 2012. I understand that this is a sum of multiple income sources across family members and negative values are generated by net losses. I have looked at the corresponding questionnaires. However, I am noticing that there is significant clumping around -9999 and I wonder if this is a result of bottom coding somewhere? Based on the documentation that I have read, there was no bottom coding after 2000. Can I assume that these are reported values?
FTOTINC is simply a summation of the INCTOT values of all family members. INCTOT is also a summation, equalling the total values of all the component income variables. So, while FTOTINC itself does not have an explicit top or bottom code, some of these component values do, which may show up in the FTOTINC distribution. The ACS TOTINC values are bottom coded at -19998, which does not explain the -9999 clustering. More likely, these clusters are resulting from single earner households with bottom coded INCBUS00 or INCINVST values.