A major task of economics and policy economics is to explain the pattern of government behavior in its interventions in the economy. This paper introduces the use of the Decision Tree methodology to help capture policy making behavior and provide extended insights into the decision making process. Decision Tree is a flexible, non-parametric, data-driven procedure which has been in use since the early 1970's, both as a classification and as a prediction tool; its use in economics has so far been rare, and in modeling regulatory agency behavior in decision-making has been virtually non-existent. Use of this approach not only reveals the significant conditions driving a phenomenon, but also provides a ranking of the conditions - information not readily available from the regression methods (logit, probit, discriminant analysis) currently used for discrete choice issues. Given a resource- constrained environment, this is expected to be a significant insight with respect to allocation of scarce resources. Variable importance as well as the threshold values at which they are informative is revealed by this procedure. EPA's rulings on pesticides which came up for Special Review between 1975 and 1995 are analyzed a decision tree methodology (CART). The Federal Insecticide, Rodenticide and Fungicide Act, which regulates all pesticide registrations and use in the USA, mandates a cost-benefit evaluation for every decision made regarding pesticide use. Essential inputs that therefore go into EPA's decision making process are cost associated with canceling a pesticide use, benefits derived from continuing with a pesticide use and risks associated with the pesticide use. This study uses this information as well as participation by various interest groups, the ruling political ideology, and other political variables to examine whether decisions made were influenced by factors other than cost benefit considerations and how these influences occurred. Results show that both risks and benefits are significant determinants of the decision choice. All the political variables: participation by interest groups, EPA administrator, political party in power and president in power are significant - pointing to the relevance of interest group influence as well as ideological influence. The results further provide very interesting and useful insights from a policy standpoint. The analysis maps out the interactions between the factors leading to a decision (cancel or continue) and indicates the threshold levels at which these factors assume importance. The analysis also shows how these threshold levels change given that other factors are in effect. EPA's policy strategy seems to follow two distinct pathways depending on the participation or otherwise by environmental groups -when environmentalists enter the rule making process, the principal influencing factors are benefits and proxies thereof. Risks seem to appear to be inconsequential. However, if environmentalists do not enter the decision making process, risks and benefits as well as political ideology become significant deciding factors. Overall, the cancellation rate is 60 percent when environmentalists participate and 22 percent otherwise. The major policy implications are that given the importance of the participation variables and risk estimates, greater resources and attention needs to be focussed on better managing these two aspects of the rule-making process.