Sentiment-induced regime switching in density forecasts of emerging markets’ exchange rates. Calibrated simulation trumps estimated autoregression

Our contribution to existing research is that we propose a novel method to generate density forecasts of foreign exchange rates using Monte Carlo simulation with regime-switching depending on global financial markets’ sentiment. The proposed approach has been examined in a one- -month ahead forecasting exercise for 22 emerging market currency rates vs. the US dollar. The key findings of our paper are as follows. We show that: (1) our forecasting method is properly calibrated based on a variety of tests and is also suitable for Value-at-Risk analysis; (2) according to the log predictive density score density forecasts produced with our method are superior to random walk forecasts in the case of all 22 analysed currency pairs, and for 7 exchange rates this advantage is statistically significant; (3) in the case of 19 analysed currency pairs our method performs better than the threshold autoregressive model (TAR) with market sentiment as the threshold variable, and for 11 exchange rates this forecasting edge is statistically significant; (4) in the case of 15 analysed currency pairs the proposed approach yields better results than the AR(1)-GARCH(1,1) benchmark, but in none of the cases this difference is statistically significant. The conducted evaluation of the proposed approach suggests that such tool can be suitable for economists, risk managers, econometricians, or policy makers focused on producing accurate density forecasts of foreign exchange rates