Dynamic Panel Probit Models for Current Account Reversals and Their Efficient Estimation

We use panel probit models with unobserved heterogeneity and serially correlated errors in order to analyze the determinants and the dynamics of current-account reversals for a panel of developing and emerging countries. The likelihood evaluation of these models requires high-dimensional integration for which we use a generic procedure known as Efficient Importance Sampling (EIS). It allows the ML estimation of panel probit models with various dynamic specifications for the error components. Our empirical results suggest that current account balance, terms of trades, foreign reserves and concessional debt are important determinants of the probability of current-account reversal. Furthermore we find under all specifications evidence for serially correlated error components and weak evidence for state dependence