Applied Methods for Forecasting Economic Aggregates and Their Components

Abstract

This thesis focuses on improving the accuracy of forecasts for economic aggregates by developing applied methods that take advantage of the strengths from both the direct and bottom-up approaches. The starting point for developing each one of these is the idea that forecasting methods that may not in themselves provide an adequate answer in a particular setting can nevertheless contribute valuable information to the forecasting process. The challenge lies in identifying and appropriately incorporating the relevant information. The first of the three methods focuses on increasing overall forecasting accuracy by jointly combining forecasts for an aggregate, any sub-aggregations and the components, from any number of models and measurement approaches. The framework seeks to benefit from each of the forecasting approaches, by accounting for their reliability in the combination process and exploiting the constraints that the aggregation structure imposes on the set of forecasts as a whole. The second method presented is one that forecasts economic aggregates using purpose-built groupings of components. The objective of developing such a method is to increase forecasting accuracy by transforming the data in a way that avoids the problems associated with disaggregate misspecification, while still allowing for distinct disaggregate dynamics to be picked up in the process. Finally, a method is developed to produce an aggregate density forecast from the density forecasts of its components in a way that considers the interaction between them. The motivation for doing this rests on the assumption that accounting for interdependencies should provide a more complete probabilistic assessment. Overall, this research shows that there are benefits to be obtained from using alternative aggregation approaches. The gains from using these methods, however, depend critically both on how they are specified and on the particular dataset. In this context, combining many specifications appears as a way of obtaining consistently good results.