At the same time, the concern that governance is significantly related to performance because it is associated with unobserved bank characteristics is important in the context of our study. In fact, the existence of such a relation is the only way to explain the results we find. In other words, shareholder-friendly boards created more value for shareholders through their decisions before the crisis, but during the crisis these decisions were associated with poor outcomes that could not be forecasted. For this explanation to work, these risks must not have been captured by traditional measures because accounting for these measures does not eliminate the relation between governance and performance we document. An example that could explain what we find is that banks with more shareholder-friendly boards invested more aggressively in highly-rated tranches of subprime securitizations. Such investments did not appear risky in 2006 by traditional risk measures, but they did work out poorly for the banks that made them. An alternative explanation for our results is that certain banks optimally chose more shareholder-friendly governance before the crisis because they were exposed to risks that required more independent board monitoring. With this view, the risks were not chosen by the board but instead led to the choice of a shareholder-friendly board. These risks had adverse realizations during the crisis, but because the banks had a shareholder-friendly board, they performed better than they would have had otherwise. With this explanation, banks with good governance had poor returns because of the risks they had, but they would have had even lower returns had they had worse governance. Governance is negatively related to performance in this case because it is correlated with risks that had adverse realizations, but it led to better performance nevertheless. Though we find some support for the latter explanation, neither explanation is consistent with the view that poor bank governance was a first-order cause of the crisis.