Forecasting with Macro-Finance Models: Applications to United States and New Zealand

In this thesis, I use macro-finance models to explore the inter-relationships between the macroeconomy and the yield curve in a forecasting setting. Using the arbitrage-free Nelson-Siegel approach to model the yield curve combined with Vector Autoregression (VAR), I jointly model macroeconomic variables and the yield curve factors to produce forecasts of inflation, activity, and interest rates. In line with earlier literature I compare whether the macro-finance model is able to better capture the dynamics of the macro variables and the yield curve factors compared with a macro-only model and a yields-only model respectively. However, a key difference is I use a full real-time forecasting setting, whereas the recent literature focuses on quasi real-time forecasting. I find there is benefit from using macro-finance models for forecasting macroeconomic variables in real-time but the gain is more significant at longer-term horizons. Indeed, the macro-finance models do not outperform traditional macroeconomic models for forecasting activity at short-term horizons. The forecasting gain is more robust for inflation and the policy rate. The theoretically motivated restrictions on the yield curve dynamics improve the forecast performance of yield curve components and generally macroeconomic variables. Using a quasi real-time environment to assess the forecast performance can overstate the usefulness of macro-finance models and understate the usefulness of placing restrictions on the yield curve dynamics.