After asserting that public institutions should not provide training for nonexistent jobs, this paper reviews problems associated with the accurate prediction of future manpower needs. The paper reviews the processes currently used to project labor force needs and notes the difficulty of accurately forecasting labor market "surprises," such as local plant relocations or shutdowns. The paper argues that "surprise-free" forecasts based on an examination of overall industry trends should be used to project normal manpower needs, while the training requirements brought about by "surprises" need to be handled in a different manner. The techniques used by the Department of Labor (DOL) to calculate manpower projections are then reviewed, their potential sources of error are identified, and recommendations for improvement are made. Next, the North Carolina New and Expanding Industries Program is described, by which the state's community colleges anticipate the "surprise" of new plant openings and subsequently train workers to begin employment there. DOL employment statistics are then cited to refute the myth that advanced computer technology will result in a manpower surplus. Finally, the paper discusses accelerated manpower demands since 1947, the demographic factors that enabled the country to meet the demands, and the flaws in predictions about the impact of computer technology. (JP)