An important analytical obstacle is the sample selection problem. Since non-employment levels are high and workers are unlikely to represent a random sample from the population of former recipients, estimates that fail to account for sample selection could be seriously biased

Previous work has documented that speaking one’s native language with an accent distinct from the mainstream is associated with lower wages. In this study, we seek to estimate the causal effect of speaking with a distinctive regional accent, disentangling the effect of the accent from that of...