Refining the Concept and Measurement of Design Effects

Abstract

The use and interpretation of design effects is hampered by variation in conceptualisation, terminology and estimation methods. This paper discusses the issues, proposes partial solutions and illustrates the ideas with examples from the European Social Survey. Three uses of design effects are identified, each with different implications for appropriate definitions and appropriate estimation methods: for sample design, for survey data analysis, and for adjustment of standard errors from standard statistical software. (A fourth use, model misspecification, is dealt with by Skinner, 1986.)