Analyses were performed to determine if the reporting sample was representative of the United States along demographic dimensions. Two samples were actually used in the development of juvenile court estimates. The first contained aggregate case counts of the number of male and female, petitioned and nonpetitioned cases handled annually by a court. The second contained an individual case record for each case processed from courts with automated information or reporting systems. Aggregate counts of the number of cases handled in 1984 were available from 1,600 courts with jurisdiction over 61 percent of the country's youth population at risk. Detailed case information was available from 1,040 courts with jurisdiction over 44 percent of youth at risk. Several strata or clustering dimensions were identified to develop more homogeneous subsamples from which to develop national estimates. These stratification factors included county size based on the number of youth in the county's population; number of youth in each county in the 1015-, 16-, and 17-year- old age groups who were under the original jurisdiction of their juvenile court; and number of youth in each age group who were white and nonwhite. It is concluded that national estimates based on a nonprobability sample are always open to doubt and criticism because such a sample cannot generate defined confidence intervals around estimates. The relative accuracy of estimates can only be assessed by comparing them with other statistics, for example, the Federal Bureau of Investigation's Uniform Crime Reports. Some of the findings of the Juvenile Court Statistics project are discussed briefly, and ways to improve national estimates are suggested. 5 tables and 1 figure