Monthly estimates of unemployment in thousands of counties, cities, and towns by the Bureau of Labor Statistics (BLS) are not reliable. Congress requires BLS to provide local unemployment estimates for use by Federal agencies to allocate funds for jobs programs and other economic assistance to local areas. Potential error in estimating local unemployment cannot be measured. Where potential errors in statistics cannot be measured, amounts misallocated cannot be determined. Developing reliable unemployment statistics for thousands of localities is a formidable assignment, far from being accomplished. Sampling such a large number of areas to obtain monthly unemployment estimates would be very costly. Although BLS has research under way on new methods, the only alternative to sampling now is to improve the present system. Local statistics are developed through a composite technique. The composite technique produces unreliable figures principally because of the methods used to estimate the amount of local labor force outside the unemployment insurance system and break down labor market statistics to smaller areas. Breakdown of labor market statistics to small areas also lacks precision. Using population is unsatisfactory because the ratios of employment or unemployment to population are not likely to be the same in all areas. BLS has taken some actions necessary to improve its statistics. However, BLS has not developed procedures for better estimating proportions in uninsured employment and unemployment in local labor forces.