Sudden stops and output: an empirical Markov switching analysis

Sudden stops and their negative effects on GDP have recently received increased attention because quantitative easing has led to substantial capital inflows into emerging economies. We extend the empirical literature on the impact of sudden stops on GDP by proposing an alternative econometric approach which is multivariate, nonlinear and uses a novel way to identify sudden stops. We estimate a Markov switching vector autoregression with a latent variable indicating whether the economy is in a sudden stop regime. We use the maximum fraction of forecast error variance approach for partial structural identification of the vector autoregression model. Beyond confirming findings from the existing empirical literature on sudden stops, our results additionally show that (i) sudden stops are associated with regime switches (i.e., breaks in the behavior of economic variables), which have significantly negative and permanent effects on GDP; (ii) impulse responses to net capital inflow shocks are regime dependent with economies being more vulnerable to shocks during the sudden stop regime; and (iii) there were different main drivers of the output decline in historical sudden stop episodes.