We studied crop harvested yield, as recorded in national agricultural statistics, to estimate net primary production (NPP) in agricultural regions where most of the land area is sown with a few, well-studied crops. We estimated the magnitudes and interannual variations in NPP in croplands in the U.S. Midwest using crop area and yield data obtained from the U.S. National Agricultural Statistics Service (NASS). Total NPP, including estimates of the above-and belowground components, was calculated from harvested-yield data by (1) conversion from reporting units of yield of the crop product (usually in volume) to mass, (2) conversion from fresh mass to dry mass, (3) estimation of aboveground yield using crop harvest indices, defined as the ratio of economic product (e.g., grain) dry mass to plant aboveground dry mass, and (4) estimation of belowground yield as a function of aboveground biomass. This approach is applied to corn, soybean, sorghum, sunflower, oats, barley, wheat, and hay in Illinois, Indiana, Iowa, Wisconsin, Michigan, Minnesota, North Dakota, and Ohio for 1992, and in Iowa for 1982 through 1996. Many counties in the eight states had Ͼ70% coverage of these crops. In Iowa, corn and soybean accounted for Ͼ50% of the land area in most counties. County-level NPP in 1992 ranged from 4 Mg·ha Ϫ1 ·yr Ϫ1 biomass (ϫ0.5 in terms of carbon) in North Dakota, Wisconsin, and Minnesota to Ͼ17 Mg·ha Ϫ1 ·yr Ϫ1 in central Iowa, Illinois, and Ohio. Areas of highest NPP were dominated by corn and soybean cultivation. NPP for counties in Iowa varied among years by a factor of 2, with the lowest NPP in 1983 (which had an unusually wet spring), in 1988 (which was a drought year), and in 1993 (which experienced floods). A sensitivity analysis, conducted by varying harvest index and root:shoot ratio by 10-50%, indicated that the limit of accuracy of the method is ϳ1 Mg·ha Ϫ1 ·yr Ϫ1 .