Download This PaperAttributes: Selective Learning and Influence
45 Pages Posted: 22 Oct 2019 Last revised: 4 Feb 2022
Date Written: February 1, 2022
Abstract
Two players who disagree on the relevance of the attributes of a complex project evaluate it based on a select few. The optimal attribute sample is characterized in a general framework in which the correlation across attributes is modeled through a Gaussian process. Two sufficient statistics inform optimal sampling: (i) the resulting alignment between the players’ estimates, and (ii) the variability of the decision. We identify an intuitive property of the correlation structure—the nearest-attribute property—that is critical for the pattern of optimal sampling. Under such a property, all optimal attributes are relevant for some player: at most two are idiosyncratic and the rest are common. The fewer and more peripheral the common attributes and the stronger the attribute correlation, the more skewed and redundant the sample. We draw testable implications for attribute-based product evaluation and strategic selection of pilot sites.
Keywords: attribute covariance, Gaussian sample paths, nearest-attribute property, strategic sampling, Ornstein-Uhlenbeck covariance
JEL Classification: D83, D81, D72, D04
Suggested Citation: Suggested Citation