THROWING DARTS AT DOUBLE REGRESSION - AND MISSING THE TARGET

To improve estimates of the extent of racial bloc voting in U.S. elect ions for which no reliable survey data on voter choice are available, social scientists have developed a ''double regression'' approach base d on use of aggregate data at the precinct level. This methodology, in tended to compensate for differential levels of turnout among minority and nonminority eligible voters, has been used in testimony involving minority voting rights issues in scores of federal trials where evide nce about the nature of voting patterns in election contests involving both minority and nonminority candidates is legally required. Wildgen makes four claims about this ecological regression methodology: (1) t hat this regression-based technique should not be called ''regression, '' (2) that double regression should not be called ''bivariate,'' (3) that it is invariably upwardly biased in its estimates of bloc voting, and (4) that it is incorrect to claim that this methodology ''cures t he ecological fallacy.'' The first criticism is absurd on its face, an d the attempts to give it a ''sensical'' meaning lead Wildgen into oth er errors. The second assertion is a trivial definitional quibble. The third complaint is dead wrong, and the fourth is directed against a s traw man largely of Wildgen's own imagination.