The statistical accuracy of Historical Foreign Trade Sources has been stated by Federico and Tena (1991) and Tena (1985, 19991 y 1992). This article follows his works in the most suspect field: geographical distribution. We have use Latin American Coal Trade Data among 1908–1930. Most international trade, considering weight, was coal trade; meanwhile it is an ideal product to isolate geographical effects. Statistical disagreements persistence makes us to think this is not a random phenomenon. We have specified an econometric model based on distance. Results show that including geography we can understand statistical disagreements. As a consequence Latin American Sources appear reasonably accurate, considering its geographical pattern.