At issue are the methods used by Federal agencies to collect, interpret, and present official figures to the public. This paper argues that these methods introduce certain distortions into official statistics which can never be fully corrected, and other distortions that may be partially corrected. Drawing from the "criminal career" research, the present analysis offers a defensible method and estimate of drug abuse that attempts to control for certain distortions inherent in official studies. Moreover, it is suggested that a longitudinal research design can explain the chronicity and persistence of drug use, and this information has significance for public policy. 2 figures, 2 tables, 1 note, and 50 references. (Author abstract)