WEAI/AERE 2009 - Individual Paper Abstract
Title: Closing the Gap between Risk Estimation and Decision-Making: Efficient Management of Trade-Related Invasive Species Risk
Author(s): Robert P. Lieli,Department of Economics, University of Texas, Austin; Michael SPRINGBORN, Department of Environmental Science & Policy, University of California, Davis, 2104 Wickson Hall, Davis, CA 95616, mspringborn@ucdavis.edu (Photo: Solanum pseudocapsicum, Jerusalem cherry. A decorative plant that has become a weed in Australia; credit: User:Fir0002 at http://commons.wikimedia.org/wiki/File:Solanum_pseudocapsicum.jpg)
Abstract:
This paper develops and compares statistical decision-theoretic tools for estimating and balancing risks and rewards, such as invasive species risk associated with burgeoning international horticultural trade. We propose a comparison of two classical methods which approach risk estimation independently from decision-making with a third technique, recently developed, which integrates this process into a single step. Under either a maximum likelihood (ML) estimation or Bayesian estimation approach, conditional probabilities of invasiveness are estimated in isolation before consequences of outcomes are considered in making the decision of whether to ban or allow a novel plant import. In contrast, in the "maximum utility" (MU) estimation approach, expected consequences have a direct influence on the estimation. The method exploits the idea that, for prediction of a binary variable (invasive/non-invasive), a global fit of the model is less important than a localized fit which partitions the information space in a way that minimizes the economic cost of classification errors. This research constitutes the first side by side examination of the MU, ML and Bayesian classification methods. To assess relative economic performance, we present an empirical application using data from an Australian import screening program for weeds. This application allows for the full exploitation of the MU framework to incorporate covariate-dependent payoffs. Initial findings suggest that this flexible utility framework is an important driver of the improvements generated by the MU methodology.