An efficiency study was conducted of the base year design used for the National Longitudinal Study of the High School Class of 1972 (NLS). Finding the optimal design involved a search for the numbers of sample schools and students that would maximize the variance at a given cost. Twenty-one variables describing students' plans, attitudes, abilities, and self-concept were selected, and 17 sets of variance components were estimated for each variable. The efficiency of these 357 statistics for each of nine designs of approximately equal cost was computed, to yield an optimal design for all statistics. Optimum designs--those with the highest average cost efficiencies--were calculated for all statistics, national statistics, and domain statistics (statistics analyzed by sex, race, and father's education). The optimal design for national statistics involves 21 students sampled per school; optimal designs for domain statistics involve 24 white students sampled per school and 30 for blacks, other races, males, females, and father's education. The current NLS design, with 18 students per school, is not far from the optimum. Designs sampling 18 to 30 students per school all have high average efficiencies for both national and domain statistics. (CP)