Macroeconomic Forecasting and Business Cycle Analysis with Nonlinear Models

The Great Recession of 2008/09 led to a sharp and persistent decline in real output growth in many countries, resulting in a global economic crisis. This economic crisis had its worst slump in economic activity in late 2008 and early 2009. At this time economists failed to predict the economic slump, though some economic indicators have shown first stirrings for the sharp downturn and thus, a severe recession. This failure raises two important questions: First, did economists use improper forecasting models? For instance, did these forecasting models not include information about the current state of the business cycle? Or did these forecasting models not account for the nonlinear dynamics inherent in the business cycle due to expansions and recessions? Second, did economists select the most informative economic indicators for their forecasting models at that time? Not all economic indicators are available at the same frequency and many models can not handle the frequency mismatch in the data. Furthermore, not all recessions are of the same origins, e.g. they might stem from financial, demand or supply shocks. Hence, the importance of economic indicators for predicting recessions might change over time and economist need an appropriate variable selection method. I examine in three independent articles different approaches to address these aspects and the problems associated with them in a real-time data analysis.