You and I both know that suburban development is encroaching on productive agricultural land in the United States, because we have the evidence of very own eyes. Each of us can think of former fields that have been converted to nonagricultural use. But how representative are observations? Advocates of growth control argue that this country is losing at an alarming and unacceptable rate (AFT 1997; Sorensen, Greene, and Russ 1997; Sierra Club 1998, 1999, 2000). Should we share their concern? We know that the use of cropland changes from year to year and that in any given year a significant fraction of cropland, perhaps one-fifth to one-third, does not actually produce any crops at all. Drought, flood, storm, diseases, insects, and other destructive forces of nature can nullify the best-laid plans of farmers, but complete crop failure is only part, and usually quite a small part, of the explanation. Good farmers rotate their crops, and many rotations require that in some year of the rotation each piece of ground lie fallow, or be used only for pasture, or grow only cover crops that are not harvested. The inherent capability of soil and climate to produce crops is only one of many factors that influence the way farmers use cropland. Land may also lie uncultivated temporarily for idiosyncratic, nonagronomic reasons: family squabbles over inheritance or divorce, compliance with or circumvention of government farm programs, quirks of wealthy landowners who do not have to extract income from every acre they possess. The use of cropland is constantly changing, and changes from year to year or from census to census may be only short-term fluctuations that are little more than statistical noise. A clear picture of cropland change in the United States requires an analysis of long-term trends that places the distractions of short-term noise in proper perspective. The dawn of a new century seems an appropriate time for a close examination of the geography of cropland change in the preceding half-century. Fifty years is time enough to provide a balanced view of long-term trends and patterns and a more robust and defensible for judging whether suburban encroachment and development threaten the future of agricultural land base (AFT 1997, xi). THE NATIONAL AGRICULTURAL LANDS STUDY The first major false alarm about our vanishing farmland was sounded in 1979, when the National Agricultural Lands Study (NALs) made scare headlines by asking, Will there be Florida oranges, fresh milk from St. Johnsbury--and Michigan tart cherries? Florida--producer of more than half of the world's grapefruit and one-fourth of its oranges--will lose virtually all of its unique and prime farm land by the turn of the century if present land loss trends continue (Fields 1979, 14). Serious students of land use immediately recognized that the NALS estimates were grossly exaggerated (Fischel 1982; Raup 1982). The NALS had derived these estimates from a comparison of two sample surveys made by the U.S. Soil Conservation Service (scs) in 1967 and in 1975. The comparison was made even though the 1975 survey was conducted in great haste, had different objectives, used different procedures and a far skimpier sample, and was never intended to be comparable with the 1967 survey. In 1984 the U.S. Department of Agriculture essentially disavowed the NALS estimates (Gustafson and Bills 1984), but these inflated estimates still circulate, and some people persist in citing them (AFT 1997, xi). Florida illustrates the errors that an undersized sample can produce. James R. Anderson, head of the Office of Geographic Research at the U.S. Geological Survey, ever the cautious bureaucrat, was so concerned about the errors in the NALS data that he dispatched Richard Kleckner to the scs office at the Iowa State University Statistics Laboratory to compare the NALS estimates with other data on a county-by-county basis in an attempt to find the source of error. …