Value of imperfect input information in agricultural production

Abstract

The paper develops a general analytical framework for evaluating imperfect information concerning a stochastic input when choosing the level of a purchased input. Second order approximations of utility yield intuitive results summarized in three propositions. Perfect information is more valuable the more it increases the efficiency of purchased input use and the more variable the stochastic input. Imperfect information is more valuable the more it increases the efficiency of purchased input use, the more the imperfect signal is correlated with the stochastic input, and the less variable the signal. A Monte Carlo evaluation of soil nitrogen testing finds that risk aversion increases the value of perfect information, while risk aversion and signal noise reduce the value of imperfect information. The propositions substantially underestimate the value of information—predicting less than half the actual value.