Office of Survey Methods Research

Abstract

Susan King, John Schilp and Erik Bergmann "Assigning PSUs to a Stratification PSU"

After every decennial census, many surveys including the Consumer Price Index (CPI) and the Consumer Expenditure Survey (CE) redefine their primary sampling units (PSUs), which are sets of contiguous counties. Since the CE survey is used to weight the CPI, the two surveys use a common set of PSUs. There are two types of PSUs: self representing and non-self representing PSUs. Self representing PSUs are selected with certainty, whereas non-self representing PSUs are grouped into a stratification PSU and one PSU is randomly selected to represent the stratification PSU. To minimize survey variance, the stratification PSUs should be homogeneous and have approximately equal populations. This is a constrained clustering problem and is solved using heuristic algorithms.