Control totals are often used in calculating weights for survey respondents. The control totals are used to calculate a factor which is used to adjust the weights. This has the effect of making the weighted total equal to the control total. The Census Bureau’s American Community Survey (ACS) and Census 2000 Supplementary Survey (C2SS) use housing unit (HU) control totals to make the weighted total number of Hus equal to the number of HUs counted in the 2000 Census. 

The ACS is a monthly survey that collects demographic and socioeconomic data about households and persons and is intended to replace the decennial census long form. Testing of the ACS began in 1996, and is now in 36 counties grouped into 31 sites. Full implementation of the ACS is scheduled to begin in every county in the United States in 2003. The C2SS used the same methods and instrument as the ACS, surveying approximately 700,000 HUs in 1,203 counties. It was conducted to test the feasibility of conducting a large scale national survey concurrently with the decennial census. The C2SS was designed to be used in conjunction with the ACS to produce national and state level estimates.

The 3,142 counties in the United States were grouped into primary sampling units (PSUs). The PSUs are grouped into strata. A strata may consist of only one PSU, and counties with more than 250,000 people are in a PSU and strata by themselves. Each ACS site is a strata by itself. Strata are classified as self representing (SR) and non-self representing (NSR). All PSUs in SR strata are selected to be in sample. In NSR strata, two PSUs are selected to be in sample. ACS sites are considered SR strata. In weighting, the strata are used to form cells for which adjustment factors are computed.

In 2000, the ACS/C2SS used the count of HUs in the 2000 decennial census for control totals. But in future years, we will not have the advantage of a census conducted in the same year to provide control totals. This paper examines the effect that HU controls have on estimates in the ACS/C2SS. W e attempt to determine if HU controls are necessary for producing accurate and reliable estimates. We also examine alternative HU estimates to determine their effectiveness as control totals.

We also examine the use of the principal person factor (PPF) in the weighting of housing units to determine if it is necessary to continue using it.

Housing controls and the PPF are used to compute adjustment factors for the HU weights after the application of nonresponse adjustment factors. The housing controls are used to compute a housing post stratification factor (HPF1) to make the weighted number of HUs equal to the control total. The HU weights are then assigned to each person in the H U. Population controls, from the Census Bureau’s population estimates program, are used to compute post stratification factors for each person on the basis of race, age, sex, and Hispanic origin. The post stratification factor of the principal person (the PPF) is then applied to the HU weight. The housing controls are then used to compute a second housing post stratification factor (HPF2) to make the weighted number of HUs again equal to the control total. The final weight is the product of the base weight (determined by the probability of selection) and all of the adjustment factors.