Rama Cont Imperial College A necessary step for the design of effective regulations and the improvement of risk management models is the causal analysis of market events which have led to risk management failures and large financial losses in financial institutions. We propose a quantitative framework for the forensic analysis of such episodes of market turbulence. We argue that a key ingredient of such a framework is the market impact of institutional investors trading activity, which may leads to nonlinear feedback and endogenous risk: many 'black swans' - extreme market moves often considered as statistical outliers- are in fact manifestations of endogenous market instabilities that arise from (non-random) feedback loops which couple price behavior and institutional order flow. We illustrate these points using data from the Quant Crash of August 2007. Based on joint work with Lakshithe Wagalath.