Self-employment and co-villager networks of internal migrants in China

This paper studies the impact of co-villager networks on the self-employment propensity of internal migrants in China. Employing a nationwide survey data, we document the self-employment patterns among internal migrants across different industries, home provinces, and individual characteristics. Applying a spatial autoregressive model and a spatial two-stage least squares procedure, we find that internal migrants are more likely to become own-account self-employed when more co-villagers in their networks are own-account self-employed. Moreover, the network effect is stronger among individuals who socialize more frequently with co-villager relatives and friends, as well as those who have lived in the host city for a longer period. We further adopt the Causal Forest with Instrumental Variable methodology to thoroughly explore heterogeneity in network effects across various feature dimensions. The results highlight that network size is a key determinant of network influence, followed by spatial and structural factors such as home province, host city, industry, and year. In addition, we find that networks that facilitate professional knowledge transfer from experienced co-villagers tend to exhibit stronger network effects, whereas those characterized by high levels of competition exhibit weaker effects.