Friday, November 8, 2013

West End Ballroom A (Washington Marriott)

*Names in bold indicate Presenter

Regulatory agencies face difficult tradeoffs when targeting facilities for environmental inspections. A rational inspection strategy is one that targets facilities to maximize overall compliance with environmental regulations while operating within the agency’s constrained resources. Such strategies are critical because agencies often have enough resources to inspect only a fraction of regulated facilities or to detect only a portion of violations. When developing an inspection strategy, a regulatory agency must establish a method to prioritize facilities for inspection, determine the frequency at which to inspect facilities and set standards for the thoroughness with which to inspect facilities. Agencies often make these decisions based on anecdotal evidence, rules-of-thumb or qualitative models, yet research has demonstrated that quantitative models that target facilities based on risk can lead to greater overall compliance than

*ad hoc*methods for a given level of agency resources. Developing such models, however, is frequently complicated by a lack of data on facility risk factors, the non-generalizability of the data that do exist, the effort required to develop a quantitative model and the difficulties associated with translating a quantitative model into a feasible inspection strategy. This paper assesses the efficiency gains obtained from various specifications of a risk-targeting approach and explores the tradeoff between amount of effort spent on data-collecting inspections and amount of effort spent on risk-based inspections. The model includes the following elements: (1) a sampling rule that splits the data into a statistical sample (for inspections intended primarily to collect data) and a risk-based sample (for inspections intended primarily to target high-risk facilities); (2) an Ordinary Least Squares regression model based on the statistical sample that predicts the amount of time required to inspect each facility; (3) a count regression model based on the statistical sample that predicts the number of violations that will be detected at each facility; and (4) a sorting algorithm that ranks facilities in the risk-based sample according to the expected number of violations per hour of inspection (as predicted from the two regression models). Using Monte Carlo simulation, the model is iterated to explore several scenarios that vary in terms of number of inspectors, the types of inspection conducted and the proportion of inspections allocated to the statistical sample versus the risk-based sample. The results are discussed in terms of the efficiency gains from various specifications of the quantitative model and the optimal tradeoff between effort spent collecting data to identify the riskiest facilities and the effort spent using these data to inspect the riskiest facilities.