American recessions since 1990 have led to recoveries that have been characterized by significantly slower employment growth than previous recoveries. Over a similar time period, there has been a documented decrease in employment in occupations that involve a high level of routine tasks due to the tendency of recent technological change to automate routine tasks. I investigate if declining demand for routine labor is driving the joblessness of recent American recoveries. I develop a search-and-matching model with routine and non-routine labor sectors wherein a negative productivity shock leads to larger employment losses for routine labor than non-routine labor, larger initial employment losses in aggregate employment, and ambiguous differences in employment growth thereafter. I use data on routine task intensity by occupation and state-level panel data of 711 detailed occupations to measure the share of each state's employment in routine occupations. In order to estimate the effect that a state's preceding routine labor share might have on the joblessness of that state's recovery, I use fixed effects regressions to exploit the variation in routine labor share and the variation in states' changes in their employment-population ratios. My results suggest that routine labor share does not explain lower employment growth during the recovery, but higher routine labor shares lead to larger per capita employment losses during the recession itself. A 1 percentage point increase in routine labor share leads to a 0.06 percentage point decrease in the change in a state's employment-population ratio over the 4 quarters leading to a recession trough. These findings have important implications for structural change in American labor markets resulting from routine-biased technological change and the effects that structural change may have on labor dynamics over the business cycle.