This paper reports evidence from a city-wide field experiment on trust. About 1000 inhabitants of Zurich take part in a trust experiment, in which first movers can condition their investments on the residential districts of second movers. First movers differentiate their investments systematically depending on where in Zurich the second mover lives. The observed discrimination pattern is robust as indicated by additional data collected in a newspaper study and a laboratory experiment. Economic status seems to be key for a district's reputation: first movers invest more if second movers live in high-income districts. Investments into districts are positively correlated with the corresponding willingness to repay, which indicates that first movers correctly anticipate the relative trustworthiness of inhabitants of different districts. Furthermore, we find that people trust strangers from their own district significantly more than strangers from other districts. This in-group effect is, at least partly, driven by more accurate beliefs about the trustworthiness of in-group members.