Pesticide Productivity Bias Due to Unobserved Variables

Abstract

This research attempts to articulate the nature of two sources of bias typically present in pesticide productivity measurements. Knowledge of pesticide productivity is useful in forming pesticide regulation, hence, numerous attempts at measuring productivity have been conducted. These measurements are typically conducted in the absence of pest population, which has the potential to bias estimates. Productivity is often measured by the pesticide marginal product, and this marginal product is often calculated without information regarding the number of times pesticides are applied, which may also be a source of bias. Theoretical modeling reveals a productivity measurement bias likely exists due to unavailable pest population information, but its sign is indeterminate. Using potato data, productivity measurements conducted with and without pest population differ little. This suggests the bias, for this data, is empirically insignificant.Pesticide marginal products behave differently when the number of applications are not held constant. Modeling shows that when application frequencies are not held constant, the pesticide marginal product can be increasing or constant, suggesting they are not the type of marginal product economists usually think about. Using potato production data, productivity measurements are significantly lower when the number of applications is not held constant, implying the absence of application frequency variables are empirically important for pesticide productivity measurements.