This paper examines two new methods to generate gridded agricultural Gross Domestic Product (GDP) and compares the results with a traditional method. In the case of Brazil, these two new methods of spatial disaggregation and cross-entropy outperform the prediction of agricultural GDP from the traditional method that distributes agricultural GDP using rural population. The paper finds that the best prediction method is spatial disaggregation using a regression approach for all the key crops and contributors to agricultural GDP. However, the issue of degrees of freedom is an important limiting factor, as the approach requires sufficient subnational data. The cross-entropy method with readily available spatially distributed crop, livestock, forest, and fish allocation far outperforms the traditional method, at least in the case of Brazil, and can operate with national- and/or subnational-level data.