Modeling and forecasting exchange rates in an unified econometric framework

In this project we aim to improve existing empirical exchange rate models by accounting
for uncertainty with respect to the underlying structural representation within a flexible Bayesian non-linear time series framework. This model approach provides a joint representation of the exchange rate as well as other latent components like the output gap or trend inflation, quantities that are typically approximated through different observed
measures like the unemployment rate. In a series of well designed forecasting exercises
we investigate whether allowing for movements in the underlying set of exchange rate
fundamentals significantly improves upon predictions stemming from traditionally used
models and the random walk benchmark. Moreover, we apply and assess whether recent
Bayesian techniques based on dynamic prediction pools help to obtain more robust
predictions relative to the forecast densities obtained from the best performing single
model.

Oesterreichische Nationalbank (Jubiläumsfonds)