SOCIAL ALCHEMY IN THE COURTROOM - THE DOUBLE REGRESSION HOAX

As readers of the exchange between Grofman (1991 a, 1991 b) and Bulloc k (1 99 1) will surmise, there is an active, if small, ''vote dilution '' subculture in American social science. Members of this subculture a re at odds over methodological issues regarding the analysis of black/ white voting patterns. Much of that dispute is over what can only be c alled the ''double regression hoax'' to which federal courts have fall en prey. The hoax consists of three prongs, which assert that (1) a te chnique known commonly as ''double regression'' (two ecological regres sions) cures the ecological fallacy; (2) ''double regression'' is biva riate; and (3) ''double regression'' is regression. None of these prop ositions is true, yet defenders of ''double regression'' resist scient ific challenges to their method by taking refuge in an argument from a uthority-their method has been approved by the Supreme Court.