Experimental investigation of consumer price evaluations

We develop a procedure to collect experimental choice data for estimating consumer preferences with a special focus on consumer price evaluations. For this purpose we employ a heteroskedastic mixed logit model that measures the effect of the way prices are specified on the variance of choice. Our procedure is based on optimal design ideas from the statistics literature and on some algorithms for constructing choice designs published in marketing journals. In an empirical application on mobile phone preferences we find evidence that the way prices are specified significantly affects the variance of choice. In a simulation study we show that our design is significantly more efficient than randomly generated designs., which can be regarded as equivalent to most commonly used experimental designs in the literature.