Christian Gourieroux University of Toronto This paper explores the modeling and measurement challenges of systematic risks and contagion for failure events, with an application to hedge funds’ survival. The dependence in individual liquidation risks results either from an exogenous common factor with joint effects on the survival intensities, or from contagion phenomena which make the intensities dependent on past liquidations. In order to get tractable models for estimation and prediction purposes, we perform the analysis at a semi-aggregate level and consider the liquidation counts of several management styles. We introduce a dynamic model for multivariate count data with both lagged count values (contagion) and unobserved factors (dynamic frailty) among the regressors. The assumptions ensure that the joint process of liquidation counts and common factor is affine to facilitate nonlinear prediction at any horizon and estimation by a new method of moments. Our empirical analysis shows that the common factor, the sensitivities to this factor and the contagion scheme can be interpreted in terms of liquidity risks. The factor is related nonlinearly to rollover and margin funding liquidity risks. The sensitivities to the factor are funding liquidity risk exposures, which depend on the redemption and leverage policies of fund managers. The causal scheme captures the reinforcing spiral between funding and market liquidity risks.