摘要
Reliability of crop production data has implications for yield gap analysis, production time trends, trading and policy decisions. In this paper, we compared databases of major grain crops estimated by a pair of independent organisations in Nebraska, USA (USDA-NASS, National Agricultural Statistics Service of USDA vs NRD, Natural Resources Districts of Nebraska) and a pair of independent organisations in Argentina (MA, Ministerio de Agricultura vs. BC, Bolsa de Cereales de Buenos Aires). The comparisons involved the yield of irrigated and rainfed maize and soybean reported by USDA-NASS and NRD, and the yield, acreage and production of maize, soybean and wheat reported by MA and BC. The comparison between NASS-USDA and NRD yield data included 127 paired observations for maize and 87 for soybean. For the pooled data involving irrigated and rainfed crops, the average difference in yield between the two sources was small (?, and higher yield above this threshold. For soybean, NRD returned lower yield than NASS-USDA for average yield below 3 t ha?, and higher yield above this threshold. For the pooled data comprising 13 regions and 9-0 cropping seasons per region in Argentina, differences between yield reported by MA and BC were larger and more scattered for maize than for soybean and wheat. The differences in acreage between the two sources increased with increasing acreage for soybean and wheat, and the same pattern was found for total production. Differences in production were more closely related to differences in acreage than to differences in yield, thus highlighting the need to improve the accuracy of crop acreage estimates. Disaggregation of data showed compensation between regions where positive differences (BC > MA) compensated negative differences (BC All four organisations providing cropping statistics involved experienced professionals using rigorous methods; hence comparisons did not seek to establish the “right-estimate. The conclusions from these comparisons are thus asymmetric: where the two sources show statistical agreement, we can have some confidence on the reliability of the data, but where the sources disagree, we cannot tell which one is more reliable; we can, however, highlight the mismatch and recommend caution in the use and interpretation of crop yield and production data, particularly at regional level.