Response Rates in Business Surveys: Going Beyond the Usual Performance Measure
Katherine Jenny Thompson, Broderick E. Oliver
Many ongoing programs compute response rates for usage both as performance measures and as quality indicators. There is extensive literature on the computation and analysis of response rates for demographic surveys, which are generally characterized by multi-stage designs with heterogeneous populations within selected clusters. In contrast, business surveys are characterized by single-stage designs with highly skewed populations. Consequently, business surveys in the Economic Directorate of the U.S. Census Bureau compute two flavors of response rates: the unit response rate (URR), defined as the rate of the total unweighted number of responding units to the total number of sampled units eligible for tabulation; and a total quantity response rate (TQRR), which is the weighted proportion of a key estimate reported by responding units or obtained from equivalent quality sources (Lineback and Thompson 2010). Thus, for each statistical period, a survey produces one unit response rate and several total quantity response rates one per key item. In this article, we describe how these two rates are computed, then introduce a statistical process control analysis perspective for monitoring them. We illustrate this approach with examples from ongoing economic programs conducted by the U.S. Census Bureau.
Performance measure, quality indicator, p-chart, general linear hypothesis test