This paper estimates a three-stage structural model of how foreign linkages affect innovation which in turn affects firm productivity. Using harmonized firm-level data from the World Bank Enterprise Surveys, I construct a panel dataset for 47 developing countries spanning 2003 to 2019. I distinguish between four types of foreign linkages (exports, imports, inward foreign direct investment (FDI) and the use of foreign-licensed technology), and two types of innovation (product innovation and process innovation). To mitigate concerns regarding sample selection and endogeneity biases, I employ advanced panel data methods. In the first stage, I find that being an exporter, using foreign inputs, and using foreign-licensed technology makes firms more likely to invest in R&D, relative to other firms. I also find evidence of sample selection bias which is corrected by using a two-step Heckman selection model. In the second stage, I find that while increases in the R&D intensity increase the probability of product innovation, they have no statistically significant effect on the likelihood of process innovation. In the third stage, I find that product or process innovation is associated with increases in firm-level productivity. These results remain robust across alternative measures of innovation and firm productivity.