Stated consumer behavior with D-efficient choice set designs : the case of mobile service bundles

A key aspect in the design of stated preference (SP) surveys is the determination of the matrix of attributes. Traditionally, practitioners have used orthogonal and similar methods to optimize these design matrices in linear and nonlinear settings. Recent research, however, has raised several concerns as to whether these methods produce optimal forecasts for nonlinear models. Because of these concerns, a different method known as efficient or optimal design was developed. This study applies D-efficient design to a nonlinear model and develops a test to measure if, and under what circumstances, D-optimization promises more accurate forecasts than traditional methods.