Despite the widespread use of Benefit Cost Analysis (BCA) in federal and state rulemaking procedures, there is little quantitative evidence regarding how the use of such analysis changes rulemaking outcomes. Though in theory BCA improves rulemaking outcomes, BCA can also increase rulemaking time, cost, and staffing requirements, causing complaints that such analysis ossifies government. Characterizing the effect of BCA on the rulemaking process remains a critical task for identifying what elements of BCA have the most significant impact on decision making. By using automated content analysis, I analyze over 150 rulemakings with corresponding BCAs, I assess how BCA content, in terms of methodology and approach, changes the likelihood of a rule transitioning through the rulemaking process. I also model the correlation of BCA content with changes in the substance of rules, thus showing what substantive content in a BCA is likely to be correlated with the attenuation or amplification of rule stringency from proposal to adoption.

I adopt a multi-state Markov event history model to analyze how the content and conclusions of a benefit-cost analysis are correlated with changes in the proposed rule. To clarify the role of BCA, I control for the characteristics of rules as well as the rulemaking agency. In contrast to other event history modeling techniques, a multistate model allows me to characterize a rule as a repeatable progression through a regulatory process rather than as a single transition, which enables me to model increases or decreases in rule stringency from proposal to adoption. I utilize the regulatory rulemaking archives generated as a result of the Washington State Administrative Procedures Act (APA) to investigate the role of BCA in state rulemaking, with the hypothesis that while BCA will increases the duration to rule adoption, differing methodological content significantly impacts the likelihood of a change to the rule under study. For the sample, I select only significant rules. I use natural language processing and supervised learning coding techniques (Bird et al 2009) to characterize the substantive content of the BCA in terms of the use of quantitative or qualitative data, presentation of uncertainty, and the number of decision options considered. I utilize longitudinal coding (Saldana 2009) to characterize the regulatory stringency and content of the rule from preproposal to adoption.

This research provides quantitative assessment of the impact of BCA on regulatory rulemaking, and thus, the conclusions are critical to identifying where improvements in BCA conduct within agencies can make for more efficient decisions and more effective decision aids. The conclusions demonstrate that the use of quantitative versus qualitative data has a significant correlation with a longer rulemaking duration, though the type of data utilized is not correlated with a change in the result of the rulemaking. A discussion of the role that uncertainty plays in decision outcome transition also provides empirical evidence of how providing uncertainty estimates is correlated with changes in the probability of rule changes. The results include substantive recommendations for policymakers regarding specific methods and how they are likely to impact rule outcomes.