State-of-the-art term structure models of commodity prices have serious difficulties extrapolating the prices of long-maturity futures contracts from short-dated contracts. This situation is problematic for valuing real commodity-linked assets. We estimate a nonlinear four-factor continuous time model of commodity price dynamics. The model nests many previous specifications. To estimate the model, we use crude oil prices and inventories. The inventory data and nonlinear price dynamics have a large impact on oil price forecasts. The additional factor in our model compared with current three-factor models has a significant impact on model-implied long-maturity futures prices.

State-of-the-art term structure models of commodity prices have serious difficulties extrapolating the prices of long-maturity futures contracts from short-dated contracts. This situation is problematic for valuing real commodity-linked assets. We estimate a nonlinear four-factor continuous time model of commodity price dynamics. The model nests many previous specifications. To estimate the model, we use crude oil prices and inventories. The inventory data and nonlinear price dynamics have a large impact on oil price forecasts. The additional factor in our model compared with current three-factor models has a significant impact on model-implied long-maturity futures prices.Received November 23, 2015; editorial decision September 12, 2016 by Editor Geert Bekaert.