Volatility Forecast of Construction Cost Index Using General Autoregressive Conditional Heteroskedastic Method

Abstract

The Engineering News-Record (ENR) publishes the Construction Cost Index (CCI) monthly, which is a composite index of 20-city average price of construction activities in the United States. Cost estimators use this index frequently to estimate the cost of construction projects. The CCI forecast provides contractors with more accurate bids. It also helps owners with their projects’ budgeting. Previous studies have forecasted this index via multivariate and univariate techniques. Homogeneity of variance is assumed on these techniques, however the CCI shows periods of substantial volatility. So far the standard tools used to address volatility on time series have been the autoregressive conditional heteroskedasticity (ARCH) and generalized autoregressive conditional heteroskedasticity (GARCH) models. In this study, a seasonal historical data set of ENR Construction Cost Index is analyzed in order to extract and forecast volatilities of the CCI in the short term. Results of this study show high and persistent volatility of the CCI in the cases of economic shocks. The maximum of the conditional variances over the entire sample period (January 1987 to July 2012) occurred in October 2008. The variance equation shows that both lags of residuals and the lag of GARCH term are highly significant, implying that the volatility of risk is influenced by past square residual terms as well as current variance of the series.