This paper presents new evidence on the employment effects of a large increase in agricultural minimum wages in South Africa using anonymized tax data. We add to the minimum wage literature by differentiating employment effects resulting from the destruction of existing jobs and from the slower creation of new jobs. Using data from tax years 2010/11 to 2016/17 and difference-in-difference models, our results show that employment decreased by approximately 14 percentage points following the minimum wage increase. Only 5 percentage points can be ascribed to job destruction, while the rest to slower job creation. Slower creation of new jobs is, therefore, the main channel through which minimum wages affect aggregate employment. Moreover, only 37 per cent of the intended increase was actually paid to workers, suggesting partial compliance with the legislation. Together, our results also provide an explanation of the paradoxical large disemployment and large non-compliance relationship; this is because employment was affected mainly by slower job creation and not job destruction.