This paper analyses the link between social networks and ethnic occupational niches in the manufacturing sector in South Africa. To this end, it employs the methodology of Bertrand et al. (2000) to minimise the omitted variable bias induced by standard approaches investigating network effects and adopts Models (1993) concentration index to define an ethnic niche. The results indicate that 25 percent of our sample is employed in ethnic niches in the manufacturing sector but that niche employment varies markedly by language group. With regards to the effect of social networks, increasing the quality or quantity of an individuals contacts by one standard deviation increases his probability of niche employment by 4 percent. Put differently, social networks magnify a policy shock affecting employment in ethnic niches by over 100 percent. This paper therefore highlights the importance of social networks, which channel workers into jobs that become ethnic niches, in the manufacturing sector in South Africa.