Our approach consists in modeling the conditional mean and the conditional variance of the exchange rate devaluation. The first and the second order moments of exchange rate devaluation are driven by the same Markov process governed by an unobservable state variable. The underlying assumption is quite simple: low mean values of exchange rate devaluation are associated with low volatility of the exchange rate devaluation, and high devaluation is associated with high volatility. We refer to the latter regime as turbulence and the former as describing ordinary market conditions. In addition, the transition probabilities in the Markov process are functions of macroeconomic and/or financial variables. This captures the idea that large exchange rate devaluations and high risk might be driven by exogenous variables.