Estimating employment responses to South Africa's Employment Tax Incentive

We present new evidence on the effects of South Africa's Employment Tax Incentive (ETI), a hiring and employment wage subsidy aimed at reducing youth unemployment. We show that attempts to estimate firm-level treatment effects via conditional difference-in-differences are likely to fail when comparing ETI to matched non-ETI firms. We show that even when eventstudy conditional pre-trends appear flat, the sensitivity of these estimates to the matching period means that pre-trends are not informative about counterfactual post-treatment parallel trends, and a broad array of matching approaches do not create credible post-treatment counterfactuals. We argue that this is likely due to mean reversion among matched non-ETI firms. A partial identification approach based on difference-in-differences with parametric time trends suggests that the ETI has increased firm-level youth employment, though some important caveats apply. Our results prompt a re-evaluation of the (sometimes contradictory) existing literature on the employment effects of the ETI: we judge that, in light of our findings, there is insufficient evidence to conclude on its impact either way.