Highway Construction Cost Forecasting Using Vector Error Correction Models

Abstract

Highway construction costs are subject to significant variations over time. These fluctuations are apparent in trends of indices, such as the national highway construction cost index (NHCCI). These variations are problematic for highway contractors because they can result in bid loss or profit loss. They are also problematic for owner organizations, such as State Departments of Transportation (State DOTs), because they can result in hidden price contingencies, delayed or cancelled projects, inconsistency in budgets, and unsteady flow of projects. These problems can be prevented if highway construction costs are forecasted accurately. Existing literature lacks appropriate models to accurately forecast NHCCI. The objective of this research is to identify the leading indicators of NHCCI (i.e., the explanatory variables of NHCCI that are useful to predict NHCCI), and create appropriate multivariate time series models for forecasting NHCCI through utilizing information available from the identified leading indicators. A pool of 16 candidate (potential) leading indicators of NHCCI, such as crude oil price, number of housing starts, and consumer price index is initially selected based on a comprehensive literature review. The leading indicators of NHCCI are identified from the pool of candidate leading indicators using statistical tests. Based on the unit root tests and Granger causality tests, crude oil price and average hourly earnings in the construction industry are found to be the leading indicators of NHCCI. Based on the results of cointegration tests, Vector Error Correction (VEC) models are created as proper multivariate time series models to forecast NHCCI. The results show that the VEC model including NHCCI and crude oil price pass diagnostic tests. The results also show that multivariate time series models are more accurate than univariate models for forecasting NHCCI in terms of out-of-sample mean absolute prediction error and out-of-sample mean square error. These findings contribute to the body of knowledge in NHCCI forecasting by rigorous identification of the leading indicators of NHCCI and creation of multivariate time series models that are more accurate than the univariate time series models for forecasting NHCCI. It is expected that the proposed forecasting models enhance the theory and practice of highway construction cost forecasting and help cost engineers and capital planners prepare more accurate bids, cost estimates, and budgets for highway projects.