Reengineering revealed preference models: international replications

Abstract

Purpose: to ascertain the validity or otherwise of Stated Preference (SP) choice models in the context of introduction of genetically modified (GM) foods with defined consumer benefits, and to report on a new experimentally designed Revealed Preference (RP) model with superior validity.

Methodology: Combining experimental conjoint choice design and a temporary and harmless ruse commonly associated with laboratory psychology experiments avoids the usual limitations of Revealed Preference (RP) choice models, and tests the validity of the SP model. The method developed in New Zealand has been closely replicated in Sweden.

Findings: In both Sweden and New Zealand, the results from the SP choice model were highly divergent from the results of the RP choice model. Research implications: In the light of these findings, SP choice models need to be treated with extreme caution when employed for predicting consumer choice of fresh fruit. These findings may extend to other food products.

Limitations: Generalised Extreme Value (GEV) modelling to compensate for heteroskedasticity has not been applied to further identify sources of model bias.

Originality and value: This new experimentally designed RP approach will improve the predictive power of market research on food choice, and will provide a benchmark for critical evaluation of SP methods. This should prove valuable in improving future market response model forecast accuracy.