We study some recent developments and challenges in the empirics of the effects of social networks. We focus in particular on researchers' ability to make policy recommendations based on a standard linear econometric model. We examine the potential compatibility between this type of econometric model and a microeconomic theoretical approach based on fundamentals, such as preferences, technology and decision processes. We discuss sources of identification for the social multiplier as well as for the identity of the key player. We study the possibility of testing endogeneity in network formation. We analyse the use of proxy variables and their impact for the causal interpretation of the peer effect coefficients. Our analysis suggests that greater care should be taken in grounding econometric network models to sound and credible theoretical underpinnings.