Acquiring insights into infrastructure repair policy using discrete choice models

Infrastructure agencies routinely make maintenance, rehabilitation, and reconstruction (MRR) decisions to keep their assets in state of good repair or to extend their service lives. They use one of three general mechanisms to make such decisions: expert opinion, continuation of historical practices, or explicit optimization using costs and benefits data for alternative MRR actions. In using any of these decision mechanisms, the agency is guided by decision factors (i.e., the attributes of the infrastructure, the operating environment, and the action in question). With regard to the historical-practice mechanism where the agency’s MRR policy is governed by past MRR decisions, there typically exists ample data on past decisions as well as the decision factors that existed at the time of the decision and hypothetically influenced the decision. Agencies that still use this decision mechanism continue to grapple with several issues, which include the feasibility of modeling the agency’s past decisions as a function of the decision factors prevailing at the time of the decision; and the influence of the decision factors on the decision outcome (MRR choice) and the temporal stability of such influences. This paper demonstrates a proposed framework to address these questions using data associated with in-service bridge decks at a highway agency in Midwestern USA. This paper also discusses the insights gained about bridge infrastructure repair policy by assessing the sensitivity of past decisions to the decision factors. The paper also demonstrates how the framework can be used to develop MRR-choice probability distributions to guide future MRR decisions and then to estimate the funding needs for future bridge deck actions. Agencies can use the methodology presented in this paper for work decisions, training of new personnel, and long-term work planning and budgeting.