Modeling and Forecasting the Price of Asphalt Cement Using Generalized Auto Regressive Conditional Heteroscedasticity

Abstract

Significant volatility in the price of asphalt cement is one of the most important challenges for both state departments of transportation (state DOTs) and highway contractors for proper cost estimating and budgeting of their projects. The ability to model and forecast asphalt cement prices can result in more accurate cost estimation and budgeting. However, there is little knowledge about how asphalt cement price fluctuates over time. The research objective of this paper is to model and forecast the price of asphalt cement using auto regressive conditional heteroscedasticity (ARCH) and generalized auto regressive conditional heteroscedasticity (GARCH) time series forecasting model which can model and predict both conditional mean and conditional variance of a variable. After analyzing the major characteristics (i.e., autocorrelation, stationarity, seasonality) of the time series of asphalt cement price, the primary conditional mean function is created using regular time series models such as auto-regressive moving average (ARMA). Then, by analyzing the residuals of this model, the conditional volatility of the price of asphalt cement is modeled using an ARCH/GARCH model. The results indicate that the developed model can predict the price of asphalt cement with less than 1.6% error.