Current estimates of livestock and poultry numbers made by the United States Department of Agriculture are derived from farmers' voluntary reports, with the help of benchmarks that are established every 5 years when a new census is taken. After every census the Department reviews the series of estimates for the preceding 5 years and makes revisions to bring it into line with the new benchmark. Rather large revisions were required in some cases when the results of the 1950 census became available. This paper discusses some of the reasons for the discrepancies and suggests changes in estimating procedures that might avoid errors of the kind made between 1945 and 1950. The suggestions are based upon the analysis of Michigan data made by Mr. Blood and Mr. Borum.