by Eric Bickel

How much is data worth? How much is it worth to reduce uncertainty? These are common questions addressed by many businesses. Quantifying the value of imperfect information can be challenging, but estimating the value of perfect information is often straight forward. In this paper [$], Professor Eric Bickel develops a relationship between the value of imperfect information and the value of perfect information. He shows that the ratio of the value of perfect information to imperfect information follows a squared, not a linear, relationship. For example, a test that is 60% accurate is only worth about 36% of the value of perfect information, not 60%, demonstrating the importance of information reliability, whereby the value of information decreases very rapidly as the reliability of the information decreases.

Published in 2008 in Decision Analysis. The abstract is free and the paper is available for purchase at this link.