I’m working with the 2016 ACS. I’m seeing single-person households where the person’s age is less than 18 and even some zeros. Seems pretty unlikely that there are 2 year olds out there living by themselves. Clearly there’s something about this data I don’t understand.
Could it be related to customizing the sample? I reduced the density to 33% because the full sample was too big to work with. When customizing a sample, are entire households kept or excluded, or are some people within a household kept while others in the same household are excluded? Meaning that my toddler householders are actually living with adults, but those adults were dropped from my sample?
Any thoughts on this are much appreciated!
I do not think this observation is being caused by your sample reduction. Taking the non-customized 2016 ACS data, limiting the data to only PERNUM==1, and tabulating age shows over 5,000 individuals who are under the age of 18, but “living alone”. This observation can be clarified by including the GQ (group quarters) variable. This shows that the majority of these cases are either “Group quarters–institutions” or “Other group quarters”. In other words, these individuals do not live in formal household units. In fact, all observations under the age of 15 are in one of these two non-household categories. There are a few individuals who are 15 years or older living alone in a formal household, but this is more realistic than a two year old.