This thesis consists of three self-contained essays that contribute to the literature on macroeconomic forecasting and empirical macroeconomics. The first essay establishes the importance of good starting conditions (i.e., nowcasts) and terminal conditions (i.e., steady-states or "stars") in obtaining accurate forecasts from vector autoregressive (VAR) models estimated with quarterly data. It does so by proposing the technique of relative entropy to tilt the VAR forecast both in the near term with the survey nowcast and in the long run with the survey long-run projection. Doing so leads to meaningful gains in multi-horizon forecast accuracy. The gains in accuracy are made possible because our proposal is an indirect approach to accommodating structural change and moving end points. The second essay develops a framework based on the model and density combinations that generate highly accurate point and density nowcasts of inflation at a daily frequency. We adopt a novel flexible treatment in the use of the aggregation function to combine density estimates from a range of mixed-frequency models. The framework permits dynamic model averaging via weights that are updated based on learning from past performance. Together these features allow non-Gaussian densities. The accuracy of the density and implied point nowcasts are significantly more accurate than the nowcasts from the survey of professional forecasters. The third essay develops a large-scale unobserved components model to estimate a range of macroeconomic stars (i.e., terminal points). The model is motivated by economic theory and empirical features such as time-varying parameters and stochastic volatility. The model allows for a direct link between the model-based star and long-run survey expectations, which significantly improves the precision of the model-based estimates of stars. The by-products are the time-varying estimates of the wage and price Phillips curves, passthrough between prices and wages, which provide new insights into these empirical relationships' instability in the US data.
|Date of Award||1 Oct 2021|
- University Of Strathclyde
|Supervisor||Gary Koop (Supervisor) & Julia Darby (Supervisor)|