Households w/out household heads?

I’m working with ACS data (1-year samples) from 2014-17. I’ve found a handful of observations in each year (<30) for which the serial number group does not have a household head (related=101) or a first person (pernum=1). Shouldn’t every unique household (grouped by year datanum serial) have one household head/first person? For these observations, the “related” variable is across the board, including spouse, child, partner/friend/visitor, etc. Does this seem correct?


First, the observation about a particular SERIAL not having an individual with RELATE==101 is driven by group quarters (or GQ). These are housing units that are not categorized as typical “households.” Therefore, if you look at only observations with GQ<3 (e.g., only typical households), then you will see that every SERIAL has at least one individual with RELATE==101. Second, I am actually unable to replicate the observation about PERNUM. Even without conditioning on GQ status, I see that every SERIAL has at least one individual with a PERNUM==1 in the ACS samples from 2014 through 2017.

Thanks, and my apologies. I realized over the weekend that my data set had been filtered to exclude individuals in group quarters, but also to filter out college-aged individuals not living with a mom or dad. So I suspect that filtering some of those observations left other observations without a household head. But thanks again for your quick and thorough response.