This study is a reanalysis of the external predictors of the use of negative concord in Philadelphia, using archival data from the Language Change and Variation survey. It is shown that the interpretation of the effects of the various socioeconomic measures reported by Labov (2001) was biased by their multicollinearity and by per-subject differences. A new mixed-effects model with residualized socioeconomic predictors and a per-subject random intercept shows the predictive role of all four socioeconomic measures, and the per-subject estimates are used to identify the nascent leaders of linguistic change.