Impacts of alternative selection in multi-alternative food demand modeling

Economic conclusions drawn from multi-attribute choice experiments are known to be sensitive to the attributes included in the choice set. This sensitivity has not been considered in multi-alternative choice experiments that are commonly used in food demand modeling. This study implements a split-sample design to determine if economic conclusions are sensitive to alternatives included in the choice set. We find that estimates of market shares, willingness-to-pay, and preference heterogeneity are conditional on the types of alternatives included. Replacing a less frequently chosen alternative with one that is more frequently chosen yields significantly different means and dispersions of willingness-to-pay for other alternatives. Relative choice frequencies and market shares change disproportionately for alternatives that are similar to the new alternative. We discuss a procedure to appropriately select alternatives and improve the likelihood of accurate choice experiment results.