Memory, multiple equilibria and emerging market crises

We present a new Generalized Markov Equilibrium (GME) approach to studying sudden stops and financial crises in emerging countries in the canonical small open economy model with equilibrium price-dependent collateral constraints. Our approach to characterizing and computing stochastic equilibrium dynamics is global, encompasses recursive equilibrium as a special case, yet allows for a much more flexible approach to modeling memory in such models that are known to have multiple equilibrium. We prove the existence of ergodic GME selections from the set of sequential competitive equilibrium, and show that at the same time ergodic GME selectors can replicate all the observed phases of the macro crises associated with a sudden stop (boom, collapse, spiralized recession, recovery) while still being able to capture the long-run stylized behavior of the data. We also compute stochastic equilibrium dynamics associated with stationary and nonstationary GME selections, and we find that a) the ergodic GME selectors generate stochastic dynamics that are less financially constrained with respect to stationary non-ergodic paths, b) non-stationary GME selections exhibit a great range of fluctuations in macroeconomic aggregates compared to the stationary selections. From a theoretical perspective, we prove the existence of both sequential competitive equilibrium and (minimal state space) recursive equilibrium, as well as provide a complete theory of robust recursive equilibrium comparative statics in deep parameters. Consistent with recent results in the literature, relative to the set of recursive equilibrium, we find 2 stationary equilibrium: one with high/over borrowing, the other with low/under borrowing. These equilibrium are extremal and “selffulfilling” under rational expectations. The selection among these equilibria depend on observable variables and not on sunspots.