Probabilistic Approach for Long-Run Price Projections: Case Study of Concrete and Asphalt

Practitioners increasingly use pavement management systems for determining the allocation of resources for multidecade investments. One important and uncertain input that will affect decisions in these frameworks is future changes in cost for rehabilitation and reconstruction actions. Despite this, existing paradigms overlook its consideration largely because little research to date has evaluated the performance of forecasting over extended time horizons. Therefore, the contribution of this study is the demonstration (via a case study) of the long-term fidelity of probabilistic price projections relative to current practice. Two paving materials, asphalt and concrete, are projected through a probabilistic hybrid forecasting model that convolves conventional forecasts for underlying constituent prices and a long-term price equilibrium relationship between commodities. Out-of-sample forecasts are conducted to test the performance of the proposed model in estimating future prices in terms of their (1) expectation and (2) prediction interval. Results indicate that the hybrid model performs similarly to current practice in terms of expectation while, and perhaps more importantly, providing theoretical uncertainty bounds that matched well with future volatility. The latter result suggests the probabilistic forecasting models developed could potentially augment current pavement management tools, allowing decision-makers to make more informed allocation choices.