Description:
We augment a standard New Keynesian model with a financial accelerator mechanism and show that financial frictions generate large state-dependent amplification effects. We fit the model to US data and show that, when shocks drive the model far away from the steady state, the nonlinear model produces much stronger propagation of shocks than the linearized model. We document that these amplification effects are due to endogenous variation in financial conditions and not due to other nonlinearities in the model. Motivated by these findings, we propose a regime-switching dynamic stochastic general equilibrium framework where financial frictions endogenously fluctuate between moderate (low risk) and severe (high risk), depending on the state of the economy. This framework allows for efficient estimation with many state variables and improves fit with respect to the linear model.