Over the past decade the price of urea has been quite volatile, especially after 2008. The high price volatility and the relatively slow transportation in the urea fertilizer industry make production planning and inventory management difficult. In this study, we construct a urea price forecasting model and compare its performance with Fertilizer Week, a commercial forecast. To construct forecast models, autoregressive (AR), seasonal autoregressive (SAR), and autoregressive-generalized autoregressive conditional heteroskedasticity (ARGARCH) models with/without exogenous variables such as Henry Hub natural gas, Brent oil, and U.S. corn prices are used with various rolling windows. Autoregressive model with exogenous variable (ARX) using the window size of 48 months outperforms our other models. There is no statistical difference between ARX with the window size of 48 month and Fertilizer Week even though Fertilizer Week is better based on forecasting accuracy measures. The combination model using the two models is statistically better than Fertilizer Week alone.