Should-Read Doruk Cengiz, Arindrajit Dube, Attila Lindner, and Ben Zipperer>: The effect of minimum wages on the total number of jobs: Evidence from the United States using a bunching estimator: "We estimate the total impact of the minimum wage on a ected employment by comparing the excess number of jobs just above the new minimum wage... http://www.sole-jole.org/17722.pdf
...following an increase to the reduction in the number of jobs below the minimum. Using variation in state minimum wages in the United States between 1979 and 2016, we find that, on average, the number of missing jobs paying below the new minimum during the five years following implementation closely matches the excess number of jobs paying just above minimum. This leaves the overall number of low-wage jobs essentially unchanged, while raising average earnings of workers below those thresholds.
The confidence intervals from our primary specification rule out minimum wage elasticities of total employment below -0.06, which includes estimates from the existing literature. These bunching estimates are robust to a wide set of assumption about patterns of unobserved heterogeneity such as regional differences or state-specific trends, measurement error in reported wages, and the precise definition of the wage band used in the bunching approach. Our estimates for the subset of minimum wage changes that affect a large share of workers are similar to the main estimates.
We also provide estimates for specific demographic groups that are policy-relevant or studied in the literature including: teens, women, workers without a college degree, women, and black/Hispanic workers. While the affected share of these groups vary considerably, the overall employment effect in each case is small and there is no evidence for substantial labor-labor substitution. We also do not find evidence for substitution away from routine-task intensive occupations. In contrast to the bunching-based estimates, we show that studies that estimate minimum wage effects on total employment can produce misleading inference due to spurious changes in employment higher up in the wage distribution...