Taking advantage of geographic (and time) variation in the proportion of routine occupations in the US, we study the impact of this variation on the wage rate of workers by educational group. Using individual data and a Bartik-type IV strategy, we show that not only non-college-educated workers but also, in the same proportion, workers with fewer than four years of college are negatively impacted by this routinebiased technological change. The latter skill group currently represents 30% of the US population. We show that only 10% to 20% of the impact on both educational groups is related to occupational and industrial downgrading (the composition eect) and that most of the wage impact occurs within industries and occupations, including manual service occupations. This is consistent with the displacement eect described in the theoretical literature on task-biased technological change and automation.