Creating new household weights in ACS data


I’m using ACS data (along with admin data) to generate a Supplemental Poverty Measure for Oregon. I have created new SPM resource units, which are defined somewhat differently from the ACS household/family, and I need to generate new household weights to go along. Two questions:

  1. Is it reasonable to assign a new HHWT number based on my new resource unit definition?
  2. Can I just use the same basic formula for assigning HHWT to a new household weight? Based on a glance at ACS data, it looks like each member of the household just gets assigned a household weight (HHWT) based on the household head’s person weight (PERWT).
  3. One odd thing – the formula above does not always work in a precise way – in some 1-person households, their HHWT is one point higher than their own PERWT (e.g., 276 vs 275; 33 vs 32) – why is this – a rounding issue? – and do I need to worry about it?


The values of HHWT and PERWT are based on the sampling design of the given sample. As is described in more detail on the ACS sample design page, the household weight value is based off the sampling integer (e.g., 1 in 10) and original housing unit weight assigned to the household based on household tenure (e.g., owned or rented), reason for vacancy (e.g., for sale or for rent), and the race of the householder. The person weight is calculated by multiplying a person factor by the household weight. Therefore, your method of using the householder’s weight to generate your new household weight should be approximately accurate. Finally, I do not think you’ll need to worry too much about the apparent rounding of some HHWT values.

Wonderful – thanks so much! I spent most of my time on the sample weights page, so thanks for directing me to the sample design page.