Estimating Macroeconomic Models of Financial Crises: An Endogenous Regime Switching Approach
40 Pages Posted: 28 Oct 2019
Date Written: June 20, 2019
We develop a new approach to specifying, solving, and estimating DSGE models with occasionally binding collateral constraints. The new specification of the collateral constraint that we propose assumes that the transition from the unconstrained to the constrained state is a stochastic function of the endogenous level of leverage. This specification results in an endogenous regime-switching model, which we solve with a new general perturbation method, highlighting the importance of using a second-order approximation. Next, using Bayesian full information methods, we estimate a well-known model with an occasionally binding constraint, with Mexican quarterly data since 1981, considering a comprehensive set of shocks. We find that the estimated model fits the data well, characterizing both their second moments and the 1995 Tequila crisis period, without imposing a negative correlation between the productivity and the interest rate process. It yields estimates of the regime probabilities that align well with the narrative of Mexico's financial crisis and business cycle history. We also find that the estimated parameters of the financial friction are tighter than previously assumed in the literature; a cocktail of shocks that co-move in a particularly averse manner, rather than a sequence of unusually large negative shocks, precede large sudden stops episodes like the Tequila crisis; expenditure and impatience shocks drove Mexico's economy into the Tequila crisis, while productivity and interest shocks prevailed during that particular episode.
Keywords: Financial Crises, Endogenous Regime Switching, Bayesian Estimation, Occasionally Binding Constraints
JEL Classification: G01, E3, F41, C11
Suggested Citation: Suggested Citation