Choosing now or choosing later: The effects of time delay and preference uncertainty on variety in repeated choice

Consumers routinely purchase multiple items at once for future consumption, such as groceries and vacation packages. Interestingly, research has shown that people choosing multiple items from an assortment tend to choose a greater variety of items when all items are chosen simultaneously as a group, before consumption, versus when each item is chosen, then consumed, sequentially one at a time. This phenomenon is referred to as the diversification bias, implying that it results from perceptual biases or biased decision processes. We take an alternative approach and show that the stochastic nature of preferences can explain why and when we observe the diversification bias. Consumers choosing now for future consumption experience greater preference uncertainty than consumers choosing for immediate consumption, due to temporally inflated stochastic variation in anticipated utility. Even without anticipating any temporal shifts in the mean liking for alternatives, simultaneous decision makers will experience future preference uncertainty and choose greater variety than sequential decision makers---leading to the diversification bias.; Using random utility theory, we demonstrate that increasing stochastic variation in utilities of available alternatives, while holding mean attractiveness of alternatives constant, will lead to the diversification bias. Using a heteroscedastic extreme value (HEV) choice modeling methodology that accounts for temporal differences in stochastic variation, we find strong empirical support that time delay between choice and consumption increases stochastic variation in consumers' anticipated future utility. Thus future preference uncertainty leads to the diversification bias.; Using a series of controlled experiments, with snacks as stimuli, we show that increasing the number of available alternatives and the density of more-preferred alternatives can induce high preference uncertainty and eliminate the diversification bias. We also show that a choice set with a very low density of more-preferred alternatives can induce low preference uncertainty and eliminate the diversification bias. In addition, we find that risk attitude does not moderate the size of the diversification bias, as previous research suggests. Finally, we demonstrate that one would need impractically large sample sizes to observe statistically significant differences in choice with the traditional regression-based measures commonly used in consumer behavior research.