Modeling hedge fund lifetimes: A dependent competing risks framework with latent exit types

•A dependent competing risks model in which the exit types are latent is proposed.•The two sets of coefficients are estimated even without observability on exit types.•There are differences in the effects of the covariates depending upon the exit type.•The two risks are dependent since frailties are significantly negatively correlated.•The time and the cause of hedge fund exit are predicted.

Due to the voluntary nature of hedge funds reporting to databases, hedge funds may stop reporting and exit a database not only because of failure, but also as a result of success and reaching the optimal size of assets under management. The existing hedge fund databases do not seem to provide reliable and unambiguous information on the reasons of hedge fund exits. In this paper, we consider that the causes of hedge fund exits are latent, and develop a competing risks model with two exit specific hazard functions, one for each cause of exit. We further allow both exit specific hazard functions to depend on unobserved heterogeneity terms that can be mutually dependent. In this way, we investigate the interdependence between the exit specific hazard functions, and explore their determinants. We show that even without observability on causes of exit, the two sets of coefficients, one for each cause of exit, can be estimated. We find an evidence of strong dependence between the unobserved heterogeneities. We also find that the estimated coefficients of the observed covariates are generally similar whether unobserved heterogeneities or the dependence between them is taken into account or not. However, the estimated exit specific hazard functions are significantly different, due to the standard negative duration dependence in the case of omitted heterogeneity.