This paper focuses on the U.S. agricultural trade against the remaining of the world. The dynamic ARDL model of error correction version is applied, not only investigating if there is J-curve effect in the short-run or not, but also taking a deep analysis for U.S. recession effects and exchange rate as well as income growth effects in the long run on the U.S. trade balance of agricultural commodities which mainly consists of bulk products and high-value products. Our results indicate that there is no significant J-curve effect for three cases, while the long-run effect demonstrates that the domestic currency devaluation is positively related with U.S. agricultural trade balance for bulk, high-value and combined agricultural products, though the high-value products appear the more modest effects compared to the other two. In sum, the real trade-weighted exchange rate is found to be the key determinant of U.S. agricultural trade balance in the long-term, rather than domestic or foreign income. We find that the three categories of agricultural products do indeed respond differently to exchange rate and income. For bulk and high-value products, U.S. exports are highly sensitive to exchange rate and foreign income, while U.S. imports barely respond. For combined agricultural products, on the other hand, U.S. exports respond greatly to exchange rate, and U.S. imports behave significantly with respect to both of changes in exchange rate and foreign income; besides, the 1980s recession had significant effects on U.S. trade balance while the most recent recession had great impact on U.S. imports, showing the U.S. trade with ROW partners was mainly influenced by the two times economic crisis during our sample period.