Changes to the Estimation Methodology
Historically, the American Community Survey has produced inconsistent estimates between
occupied housing units and householders as well as other estimates that should logically be
equal. For example, in 2005 there were approximately 3.7 million more householders than
households. In addition, there were 1.8 million more spouses than married couple households
and 176,000 more unmarried partners than unmarried partner households. These discrepancies
were mainly due to the independence of the housing unit and person weighting. This led to
research to revise our weighting methodology to minimize these inconsistencies.
Beginning with the 2006 ACS the estimate of householders will be equal to the estimate of
households and the estimate of occupied housing units because the housing unit weight for
occupied housing units is equal to the person weight of the householder. This weight is used for
tabulations of occupied housing units and households. The differences between spouses and
married couple households are reduced as well as the differences between unmarried partners
and unmarried partner households. The housing unit weight for vacant units is calculated based
only on housing unit characteristics and is equal to the final housing unit weight prior to the
person weighting and after the housing unit controls from the Population Estimates Program
(PEP) are applied.
The final estimate of housing units is no longer controlled to the housing unit estimate from the
PEP. Thus there will be a margin of error associated with housing unit estimates. This is seen in
several of the housing tables where the estimate of total housing units is shown. Due to the
improved estimate of the vacancy rate, there will be an expected increase in the estimate of
vacant units and an expected decrease in the estimate of occupied units, which is shown in tables
such as Occupancy Status (B25002). However, as a result of allowing the estimates of occupied
and vacant un