Economic Forecasting under Macroeconomic Uncertainty

点击翻译按钮获取中文摘要

Macroeconomic forecasts are essential inputs in economic decisions such as the decision to invest in a new business project, buy a house, and change the policy interest rates. Macroeconomic forecasts are computed based on models able to capture regularity in past data, such that past economic history help us to anticipate the future values of output growth and inflation. Forecasting models are developed based on different assumptions about the economic structure and the statistical properties of the macroeconomic data. During the 00's, structural models called Dynamic Stochastic General Equilibrium (DSGE), estimated using Bayesian Econometrics techniques, became popular in central banks as a tool to deliver both macroeconomic forecasts and policy analysis. Their popularity was justified by rigorous theoretical background - built on micro-foundations - and by their ability to forecast output and inflation one- and two-years ahead. However, Del Negro and Schorfeide (2012) show the shortcomings of these models in predicting the drop in output observed in the 2008-2009 recession. In contrast, Stock and Watson (2012) argue that a statistical model, called Dynamic Factor Model, was able to capture the severe downturn observed in the United States in 2008-2009. Additional recent forecasting models proposed by the academic literature are: Bayesian Vector Autoregressions and Mixed Data Sampling Models. However, for these state-of-art models, little is known about their capabilities of forecasting under macroeconomic uncertainty. From the point of view of the economic decision maker, it is hard to ascertain if the most adequate set of forecasting models (or a model) is being used, considering the current economic climate. The main output of this research project is a paper evaluating the relative forecasting accuracy of state-of-art macroeconomic forecasting models, including their performance during the recent global downturn.