The weights developed over this 9-step process play an important role in analyses of ACS 1-year data. They are designed to reduce the effects of sample bias on the ACS estimates, and they affect the precision of those estimates. The Census Bureau has conducted a substantial amount of useful research on many of these steps. However, with the data for the first year of the full ACS available, it is now time to carry out a thorough review of all steps in the process, individually and in combination, to determine if improvements can be made. The review will need to consider the impact of the current and alternative weighting schemes on the quality of a wide range of ACS estimates, covering a variety of characteristics across the full range of governmental units.
The review should start with the design decisions that lead to the sample sizes and initial base weights for governmental units of all sizes and the CAPI subsampling rates that depend on expected mail and CATI response rates. For example, the choice of subsampling rate depends on the relative costs of mail, CATI, and CAPI responses and can be informed by theoretical results on the optimum subsampling rates for initial nonrespondents developed by Hansen and Hurwitz (1946).
Although many of the adjustment factors at subsequent steps are not large, they nevertheless deserve a careful assessment to see if improvements can be made, and their combined effects need to be examined. The inconsistency in the logic for the VMS factor and the point-in-time PE housing unit and population control factors needs to be reviewed. The effects of the nonstandard nonresponse adjustments using the mode bias noninterview factor on biases and variances need examination. The case for the integer rounding of the weights needs to be critically assessed. Most importantly, the PE housing unit and particularly the population controls deserve special scrutiny, as discussed in the next two sections.
Recommendation 5-1: The Census Bureau should conduct an in-depth review of the weighting scheme used for producing ACS 1-year period estimates and assess a range of alternative schemes that might improve the quality of the estimates.
After step 4 in the weighting process, the weighted total number of all units sampled from the MAF for the year is equated to the number of units in the MAF from which the sample was selected. The main sample for a given year is selected in August of the previous year and a supplemental sample is then added in January to give representation to units that have