Why more can be less: An inference-based explanation for hyper-subadditivity in bundle valuation

We conceptualize, develop, and test a multiple-item bundle valuation model through which decision makers are able to make inferences about the value of uncertain items based on the value of certain items. Results of four experiments indicate that bundling a low-value certain item with a high-value uncertain item, which are not substitutes, results in a bundle valuation lower than the value of the uncertain item alone. We refer to this highly unexpected and previously unexplained phenomenon as “hyper-subadditivity.” In addition we find that bundling a high-value certain item with a low-value uncertain item leads to superadditivity, even though the items are not complements. Hence, we find that when two objects are bundled together, and one has a more certain value, decision makers use the value of the certain item to infer the value of the less certain item. They might infer that the other (less certain) object must be worth an amount similar to the item with which they are paired. We further demonstrate that reducing uncertainty eliminates these effects, and that direct value inferencing (not simple numeric priming, nor inferences about quality) is the most likely mechanism driving these effects.