What drives financial contagion? The empirical literature aimed at modeling financial risk spillovers in crisis periods and documenting the role of contagion channels is subject to an endogeneity issue, as the channel itself can respond to a change in the level of risk. We tackle this issue by using a novel spatial econometric estimation procedure based on a control function approach and offer "robust-toendogeneity " evidence on the role of indirect financial contagion channels in the banking industry. Our estimations, based on on 28 large US banks during the financial crisis (2007Q3-2013Q2), confirm that several channels are endogeneous. Accounting for endogeneity is proved to be important for recovering reliable estimates of transmission mechanisms. Banks similarity in fundamentals, similarity in investment strategy as well as common exposure appear as significant drivers of contagion. Based on relevant transmission's channels, we build a simple systemic risk indicator named "Interaction Based Centrality". We show that it may help forecast vulnerable institutions in times of crisis and could thereby be used for monitoring purposes by regulatory authorities.