Forecasting Economic Growth and Predicting Recessions: An analysis of yield spread, consumer confidence, and news sentiment

This Bachelor’s thesis provides a structured comparison of the indicators yield spread, consumer confidence and news sentiment in their usefulness to forecast economic growth and predict recessions in the United States and Germany. The variables are tested for normality, unit roots and cointegration. For economic growth forecasts, simple regressions and vector autoregression models are used. To predict recessions, Probit models are fitted. The in-sample fit is evaluated using AIC and out-of-sample forecasts are made and evaluated using RMSE. Which combination of variables performs best is highly reliant on the country of interest and the model estimated.