Government policy choice requires comparison of the costs and benefits of alternative actions. Difficulties arise from uncertainty about the consequences of candidate policies and the need to compare outcomes that differ on both economic and health dimensions. Economics can improve risk assessment and policy choice by providing methods for (a) identification and analysis of the scientific uncertainties that are critical to government policy selection, (b) analyzing market interactions that affect exposure to the hazards in question, and (c) valuing health effects and other potential consequences. In food supply, as in other domains, the link from government policy to health and economic outcomes is mediated by market interactions. Risk assessment (of chemicals, microorganisms, and other agents) considers the links from policy to human exposure to health consequences. The link from exposure to health effects is characterized by a physiological dose response function; that from policy to exposure depends on economic interactions. Conditional on government regulations, food producers decide what foods to produce and technologies to employ; consumers decide what foods to consume. These decisions lead to supply and demand functions, which jointly determine the quantities of each food consumed. The quantities and their distribution in turn determine consumer exposure to foods and associated agents, and worker or environmental exposure to agents associated with production. Supply and demand decisions can be affected by information about the benefits and costs associated with consumption or production. From a producer's perspective, one of the costs of a new production technology is the possibility of liability for unintended damage to consumer or worker health or to the environment. Fleisher analyzes producers' possible liability for environmental damage associated with the use of production technologies involving genetically engineered microorganisms (GEMs). Because GEMs are perceived to represent a low-probability/high-consequence risk, estimating future liability is particularly difficult. Individuals frequently have difficulty in evaluating small risks, choosing either to ignore them or to overvalue a reduction to the apparent certainty of no risk from the source (Kahneman, Slovic, and Tversky). In a new legal area, producers must consider the possibility of retroactive liability: even if all reasonable precautions are taken, GEM users may find themselves liable in the future. To the extent that identification of the producer whose GEMs caused any damage is difficult, however, individual producers may escape liability unless something like the market-share liability doctrine applied to asbestos-linked worker injury is applied. Economics can also provide a method for valuing diverse policy outcomes. Policy choice frequently involves risk-cost tradeoffs when improved health can be obtained at additional cost. Risk-risk tradeoffs may also be important; e.g., possibly carcinogenic preservatives may reduce the incidence of acute illess from foodborne pathogens, at the cost of increased cancer in the future. Candidate policies are most readily compared if the differences in outcomes on all relevant dimensions can be made commensurate, as by valuing them in monetary units. Roberts presents a simple rule for estimating the cost of illness from foodborne pathogens. The cost is assumed to be a linear function of severity, measured by the proportion of cases that prove fatal. The specific formula she presents requires validation using information on additional pathogens, but the approach is promising. Methods requiring only limited data can be valuable because the number of hazards that must be evaluated overwhelms current methods; e.g., two-thirds of pesticide James K. Hammitt is a mathematician with the Rand Corporation.