Attribute-value functions as global interpretations of attribute importance

In order to better understand decision maker’s perceptions of the importance of attributes, Goldstein (1990) differentiates between global and local interpretations of attribute importance. While the appreciation for the distinction is growing, research on the relationship between measures of global and local importance is inconclusive. We believe that these inconclusive findings are caused by operationalizing global attribute importance with single-point measures that implicitly assume that the global interpretation of attribute importance linearly depends on the relevant range of context-specific attribute levels. To address this, we propose to operationalize the global interpretation of attribute importance by estimating decision makers’ attribute-value functions. Two empirical studies demonstrate that the shape of attribute-value functions changes from concave to convex with global attribute importance. Furthermore, the steepness of these functions increases with global attribute importance while the diminishing sensitivity decreases. Finally, it is demonstrated that the inconclusive findings about the relationship between common, single-point measures of global and local attribute importance is driven by non-linearities in decision makers’ attribute-value functions. The results suggest great promise for future research on using decision makers’ attribute-value functions for measuring the importance of attributes.

► Differentiating between global and local attribute importance is critical.
► The nature of the relationship between global and local importance remains unclear.
► Attribute-value functions offer a conceptual explanation for this relationship.
► Predictive accuracy of attribute-value functions is significantly better than random model.