This paper describes a composed error model for estimating the conditional distribution of yield loss to serve as an insect damage function. The two-part error separates yield variability due to pest damage from other non-pest factors such as soil heterogeneity, non-uniform application of agronomic practices, and measurement errors. Various common functional forms (linear, quadratic, Cobb-Douglas, negative exponential, hyperbolic, sigmoid) for the pest damage function are presented and parameter estimation is described. As an empirical illustration, the model is used to estimate a damage function for corn rootworm, the most damaging insect pest of corn in the United States. The estimated damage function gives expected proportional yield loss as a function of the root rating difference and is used to estimate yield loss due to rotation resistant western corn rootworm in east-central Illinois. The estimated average yield loss is 11.6%, more than enough to cover the cost of a soil insecticide application. However, tremendous variability in actual loss exists, so that the probability that actual loss is less than the cost of a soil insecticide ranges 32-45%, depending on the assumed yield and price. As a result, IPM methods potentially have great value, since they can eliminate uneconomical soil insecticide applications.