Modeling the costs of bridge rehabilitation

Intermediate and long range planning for Departments of Transportation necessitates the estimation of the cost of bridge rehabilitation, a range of activities falling between maintenance and replacement. This paper presents a set of regression models designed within the context of data available to the New York State Department of Transportation. The nature of the data necessitated the use of non-linear regression models which were developed using the software package TRIO. Various approaches were explored in an effort to determine which group of models performed best. For each approach models were developed for estimating separately the cost of rehabilitating the major bridge components: deck, superstructure, and sub-structure. One approach involved estimating costs of each component for the entire bridge. Another approach involved unit costs. In addition models were developed which allowed comparison of the predictive power of using low bid vs median bid as the dependent variable. Furthermore, models were developed separately for bridges with single spans and those with multiple spans. The approach that consisted of working with the costs for the entire bridge consistently performed well. In many cases the relationship between the dependent variable (cost) and the predictor variables was highly non-linear. Due to differences in data availability and definitions the models cannot be applied directly by states other than New York. However, the procedures followed and approaches employed should provide guidance to those wishing to generate models to estimate bridge rehabilitation costs.