Forecasting Engineering News-Record construction cost index using multivariate time series models

The construction cost index (CCI), which has been published monthly in the United States by Engineering News-Record (ENR), is subject to significant variations. These variations are problematic for cost estimation, bid preparation, and investment planning. The accurate prediction of CCI can be invaluable for cost estimation and budgeting of capital projects, and can result in accurate bids. The research objective of this paper is to create appropriate multivariate time series models for forecasting CCI based on a group of explanatory variables that are identified by using Granger causality tests. The results of cointegration tests recommend vector error correction (VEC) models as the proper type of multivariate time series models to forecast CCI. Several VEC models are created and compared with existing univariate time series models for forecasting CCI. It is shown that the CCI predicted by these VEC models is more accurate than that predicted by the previously proposed univariate models (i.e., seasonal autoregressive integrated mean-average and Holt-Winters exponential smoothing). The comparisons are based on two typical error measures: mean absolute prediction error and mean squared error. The primary contribution of this research to the body of knowledge is the creation of multivariate time series models that are more accurate than the current univariate time series models for forecasting CCI. It is expected that this work will contribute to the construction engineering and management community by helping cost engineers and capital planners prepare more accurate bids, cost estimates, and budgets for capital projects.