Statistical models for the estimation of bridge replacement costs

Abstract

Making accurate estimates of bridge replacement costs is essential to assess present and future bridge funding needs. A series of analyses of variance was performed on bridge replacement costs to evaluate the effects of bridge attributes. Replacement cost prediction models were then developed by regression techniques. Bridge attributes which can be easily understood by bridge inspectors and engineers were used as predictor variables. Nonlinear and log-linear models were evaluated for developing cost prediction models. A residual analysis of these models showed that log-linear models were preferred to nonlinear models. Costs of bridges that had been replaced between 1980 and 1985 by the Indiana Department of Transportation (INDOT) were used as a data base. Replacement costs were converted to 1985 price using construction price indices. The final cost prediction models were validated using the costs of selected bridges which were replaced between January and June 1986, by the INDOT. Bridge replacement costs estimated by these models showed a fairly good correlation with the actual contract costs. To estimate current or future costs at a place other than in Indiana, one need to multiply appropriate cost indices.