Incorporating model uncertainty into the generation of efficient stated choice experiments: A model averaging approach

Abstract

Stated choice (SC) studies typically rely on the use of an underlying experimental design to construct the hypothetical choice situations shown to respondents. These designs are constructed by the analyst, with several different ways of constructing these designs having been proposed in the past. Recently, there has been a move from so-called orthogonal designs to more efficient designs. Efficient designs optimize the design such that the data will lead to more reliable parameter estimates for the model under consideration. The literature dealing with the generation of efficient designs has examined and largely solved the issue of a requirement for a prior knowledge of the parameter estimates that will be obtained post data collection. Nevertheless, problems related to the fact that the efficiency of a SC experiment is related to the variance-covariance matrix of the model to be estimated and that different econometric models will have different variance-covariance matrix, thus resulting in different levels of efficiency for the same design, has yet to be addressed. In this paper, we propose the use of a model averaging process over different econometric models to solve this problem. Via the use of a case study, we show that designs generated using the model averaging process prove robust to different model estimation as well as provide decent levels of protection against biased parameter estimates relative to designs generated specifically for a given model type.