Results of Four Methods of Sampling Individual Farms

Abstract

Since January, 1937, the Central Statistical Board, Bureau of the Census, and Bureau of Agricultural Economics, have cooperated in studies designed to test various sampling procedures that might be used in making surveys of agriculture. Assistance in this research on Sample Farm Census methodology has also been had from the Departments of Economics of Iowa State College and Kansas State College, from the Works Progress Administrations of Iowa and New York City, and from various State offices of the Bureau of Agricultural Economics. The plan for a sample census of agriculture originally advanced by Mr. J. B. Shepard contemplated the use of the Minor Civil Division as a sampling unit with certain restrictions and modifications.' The administrative advantages of the Minor Civil Division are undeniable. The Minor Civil Division is the smallest political unit for which U. S. Census data are tabulated. Since it is a definite geographical unit, its use as a sampling unit would facilitate the measurement of the movement of land into and out of agricultural production. Studies completed early in 1937 however indicate that, with the great amount of variability in agricultural data, the Minor Civil Division was too large a sampling unit to serve as a basis for county estimates. It could be used for estimates by areas, such as crop reporting districts or type of farming areas, on a few major items which are well distributed among farms. Even then, a 20 per cent sample was a minimum. The problem of determining the correct sampling procedure for a sample census of agriculture involves two fundamental considerations: first, the determination of the type of sampling unit which will, under a given set of conditions, give the greatest accuracy with the minimum of administrative difficulty and cost; and second, the determination of the size of sample necessary for a specified standard of accuracy. The expense of taking an annual sample census of agriculture would be difficult to justify unless such an enumerationsurvey would furnish the basic data necessary for making estimates on major agricultural items by counties, or type-of-farming areas. Accordingly, a project was outlined in September, 1937, for investigating the statistical possibilities of sampling units smaller than