Academic journal article
By Norwood, F. Bailey; Marra, Michele C.
Journal of Agricultural and Resource Economics , Vol. 28, No. 3
Pesticide productivity is both important and difficult to measure. Typically, pesticide marginal products are estimated without information on the pest pressure. Three theoretical models are developed which suggest absence of such information may cause an underestimation of pesticide productivity. Using application frequency variables as a proxy for pest populations, we show that pesticide marginal products are higher when pest pressure is accounted for.
Key words: damage abatement, marginal product, pesticide economics, productivity, unobserved variables
To protect public health, the government has aggressively pursued pesticide regulation through a series of laws beginning with the Federal Insecticide, Fungicide, and Rodenticide Act in 1947. Pesticide regulations have evolved such that today the only pesticides permitted are those which ensure "reasonable certainty that no harm will result from aggregate exposure to the pesticide chemical residues." This criterion makes the development and approval of new pesticides expensive, and has led to a 7% to 9% decrease in pesticide registration (Fernandez-Cornejo, Jones, and Smith).
A total ban on pesticide use in the United States has been estimated to cost $41 billion per year in higher food costs and lower quality crops and livestock (Knutson et al.). Thus, good pesticide policy clearly must consider the costs as well as the benefits of pesticide regulation. Economists often assess pesticide benefits by measuring pesticide marginal products. The higher the value of the pesticide marginal product relative to marginal cost, the greater the additional benefit from increasing pesticide use, and thus the greater the cost from more stringent regulation. If the value of the pesticide marginal product is low relative to marginal cost, it is more likely that farmers can profitably decrease pesticide use while simultaneously reducing environmental and health risks. It is clear, then, that regulatory mistakes can be made if pesticide marginal products are mis-measured. One source, among several, of mis-measurement is an inherent bias in the estimation of the marginal product due to choice of measurement procedure or data. This study examines analytically and empirically the bias due to omission of pest pressure in the estimation of pesticide marginal products.
Early attempts to measure the value of pesticide productivity found it to be quite high relative to pesticide marginal costs. Of the estimates conducted prior to 1986, 18 out of 20 suggest an extra dollar spent on pesticides generates more than a dollar in return (Headley; Campbell; Fischer; Carlson), implying pesticides are systematically underused from a profit maximization point of view. In response, two possible sources of an upward bias in estimated marginal products have been put forth in the literature. First, almost all studies to date use cross-sectional data from private farms where data reflecting differences in land quality, managerial ability, and other fixed effects are not available (Campbell). If any of these fixed effects are correlated with pesticide use, then the corresponding marginal products may be biased. Carpentier and Weaver found, when fixed-firm effects are accounted for, marginal products are indeed lower.1
Second, all marginal product estimates before 1986 use the Cobb-Douglas production function. In a 1986 analysis, Lichtenberg and Zilbennan argued that the single-equation, Cobb-Douglas model may be inappropriate, and offered a different model of the pesticide-yield relationship which might result in lower marginal products. This approach proposes a damage abatement model. Essentially, the approach considers the effect of pests on yield separately from the effect of pesticides on pests.
While some subsequent studies have found, under some circumstances, damage abatement models yield lower marginal products (e.g., Babcock, Lichtenberg, and Zilberman), others have reported higher estimates (e. …