What's worst? What's first? Conservatism, risk assessment and priorities at hazardous waste sites

The EPA has been criticized for its use of conservative estimates of risk in the Superfund program. Because of perceived inefficiencies, it has been suggested that central risk estimates be used in risk assessment, and for risk uncertainty to be quantified. The first chapter examines the impact of central risk estimates in the Superfund program. It is discovered that the impact on site risks is substantial if central estimates of risk are considered. Using program risk thresholds as a guide, there may be substantial impact on sites that mandate remediation based on risk. Some evidence for misordered priorities is also found.;The second chapter seeks to quantify the conservatism in EPA risk estimates through Monte Carlo simulation. The resulting risk distributions are used to assess alternative policy specifications based on consistent margins of safety about the risk estimates. Results indicate that EPA ingestion risks are highly conservative, with sixty percent of assessed pathways exceeding the 99th percentile. Using agency thresholds as a guide, there is a significant reduction in the number of sites that require remediation when a margin of safety using the 90th or 95th percentile of the risk distribution is employed.;While EPA currently focuses on individual risk levels, benefit-cost analysis would require that we consider the number of persons exposed to site risks. The third chapter constructs standardized, probabilistic risk for populations around Superfund sites in order to obtain a distribution of the potential cancers posed by site contamination and the cost per cancer averted at each site. Three alternative approaches to priority-setting are examined: individual risk, population risk and cost-effectiveness. Results indicate that the cost per cancer averted is in the several hundred million dollar range for most Superfund sites and that priorities based on population risks more closely mirror a cost- effective approach.