Adaptive survey designs for nonresponse and measurement error in multi-purpose surveys

Recently, survey methodology literature has put forward responsive and adaptive survey designs as means to make efficient tradeoffs between survey quality and survey costs. The designs, however, restrict quality-cost assessments to nonresponse error, while there are various design features that impact also measurement error, e.g. the survey mode, the type of questionnaire (long or condensed) and the type of reporting (self or proxy). Extension of adaptive survey design to measurement error is, however, not straightforward when a survey has many and diverse survey items. An adaptive survey design needs to make an overall choice of design features that applies to all survey items simultaneously. In this paper, we investigate adaptive survey designs that account for both nonresponse and measurement error. In order to do so, we model the underlying causes for differences in measurement error between design features. This leads to response styles or response latencies. We tailor efforts so that either response style propensities are minimized or constrained. We illustrate the ideas with a case study on the 2008 Dutch Labor Force Survey. The design features in this study are the type of reporting (self-reporting only versus proxy-reporting allowed), and the number of face-to-face calls.