Dynamic Macro Scenario Analysis via Bridge Sampling

Abstract

Dynamicmacroeconomicmodels shouldby designbe amenabletomacro scenario analyses under some stated policy objective or presumed stress environment, for example, anticipating how an economy would look like with an inflation target or under severe economic downturn. However, such analyses are difficult to conduct mainly due to the technical complexity associated with partially conditioning on a future scenario. In this paper, we devise a generic bridging sampling method for dynamic scenario analyses that is flexible with scenario setting while accommodating parameter uncertainty. In contrast to the literature, our method is efficient and can be straightforwardly applied beyond linear-Gaussian models. We demonstrate that macroeconomic models with comparable forecasting performance can have very different implications while being considered under a policy scenario. Dynamic scenario analysis thus goes beyond helping policy formation, and now provides a new angle for discriminating competing models with a specific reference to real-world usage.