Citation indexes are increasingly used to measure the scientific impact of researchers and institutions, though their application is often criticized. We study the network of citations of all publications indexed in Web of Science authored or coauthored by Italian tenured academic economists. We show that citations capture many factors other than mere scientific quality. By estimating the determinants of the probability that any author is cited by any other author, we find those factors to involve not only similarity in methods and topics but also, significantly, various measures of social community as well as of political proximity. Our analysis leads us to conclude that, at least in the case of economics, citations cannot be interpreted as mere proxies of scientific impact, and their use to produce indexes and rankings may require careful rethinking.