In this study we investigate the link between the job search channels that workers use to find employment and the probability of occupational mismatch in the new job. Our specific focus is on differences between native and immigrant workers. We use data from the German Socio-Economic Panel (SOEP) over the period 2000-2014. First, we document that referral hiring via social networks is the most frequent single channel of generating jobs in Germany; in relative terms referrals are used more frequently by immigrant workers compared to natives. Second, our data reveals that referral hiring is associated with the highest rate of occupational mismatch among all channels in Germany. We combine these findings and use them to develop a theoretical search and matching model with two ethnic groups of workers (natives and immigrants), two search channels (formal and referral hiring) and two occupations. When modeling social networks we take into account ethnic and professional homophily in the link formation. Our model predicts that immigrant workers face stronger risk of unemployment and often rely on recommendations from their friends and relatives as a channel of last resort. Furthermore, higher rates of referral hiring produce more frequent occupational mismatch of the immigrant population compared to natives. We test this prediction empirically and confirm that more intensive network hiring contributes significantly to higher rates of occupational mismatch among immigrants. Finally, we document that the gaps in the incidence of referrals and mismatch rates are reduced among second generation immigrants indicating some degree of integration in the German labour market.