What are the rationales underlying consumers' grocery purchases that typically involve multiple varieties or flavors of the same brand? Liang Guo of HKUST considers the example of yoghurt, where consumers are likely to purchase multiple items and even multiple flavors, and breaks down the underlying rationales to offer insights on what consumers may be thinking when they make their choices.

Two rationales are of interest. One is the self-detraction in the consumers' personal preferences, called "negative state dependence", whereby the consumption of one variety leads to a lower future preference for consuming the same variety, and thus, a higher preference for other varieties. The other one is the consumers' uncertainty, at the time of purchase, about what they will prefer at the actual time of consumption, encapsulated in a desire for "consumption flexibility." Both forces can drive a consumer to buy different varieties/flavors together in a single shopping basket.

"Although a positive state dependence may enhance a consumer's incentive to buy the same varieties simultaneously, a negative one may enhance their purchase of different varieties. The need for consumption flexibility may also lead to purchases of horizontally varied products. As a result, a horizontally varied assortment purchase can be driven by either negative state dependence or consumption flexibility," he says.

How can these two rationales be empirically separated when only purchase, but not consumption, information is available for the researcher? This is the challenge addressed by Guo, which is believed by previous researchers to be impossible.

Guo develops a model to measure both state dependence and consumption flexibility due to utility uncertainty. He applies the model to scanner panel data, where only purchase information is available. The data comes from several hundred households as provided by A.C. Nielsen, covering January 1986-August 1988. He carefully selects only single-person households as the calibration sample for model estimation, in order to rule out alternative explanation for the observed purchase pattern.

The model proves to be a good fit for the data. The estimation results suggest that, despite the desire to consume the same variety across consumption occasions (i.e., positive state dependence), buying different varieties together can increase the expected value of the purchased assortment because future consumption choices will become more flexible. That is, both positive state dependence and consumption flexibility can co-exist and exert influences on consumer purchase behavior. Guo also emphasizes the importance of accounting for both effects, since ignoring one effect may lead to misleading insights on the other.

The consumers in the data show a tendency to purchase one flavor, mixed berry, on its own rather than in varied assortments as seen with other flavors in the study (strawberry, cherry, blueberry and pina colada). "This suggests that consumers at the purchase time are pretty sure of their future utility for mixed berry, whereas the consumption utility of the other varieties is dependent on future contingency," Guo says.

Based on the estimated result of positive state dependence, Guo shows how free sample distribution can help managers increase demand: it increased sales return as high as 21 per cent for mixed berry, but made no real dent on strawberry, probably because consumers had a higher intrinsic preference for strawberry and as a result this was much more difficult to elevate.

The model also offers a means of measuring the magnitude and the demand impact of utility uncertainty. For instance, in the case of cherry yoghurt, an increase in uncertainty can improve sales by 33 per cent. Guo points out that consumers' uncertainty perceptions can be changed through such things as commercials.

"It seems that increasing uncertainty can improve aggregate demand only for those varieties with both relatively high uncertainty estimate and relatively low intrinsic preference. It is the purchase of these varieties that is most likely to be driven by the consumers' desire to maintain flexibility," he says.

Overall, Guo's model and tests provide strong empirical support for observed patterns of assortment choices. "This research demonstrates that the absence of consumption information may not necessarily prevent us from making behavioral inference from purchase data," he says.