University of Southampton
The measurement of expert judgement uncertainty in central bank forecasting
Abstract
dc:description.abstractThis thesis provides a comprehensive analysis of the determinants of performance and behavioural uncertainty in professional forecasters’ macroeconomic predictions. It introduces a novel framework that integrates statistical, psychological, and computational techniques to model behavioural uncertainty under rational expectations. Additionally, it makes empirical contributions by reanalysing the UK-SPF datasets, offering fresh insights into the use of survey-based projections for real-world decision-making.<br/><br/>The thesis comprises three main chapters. Chapter 2 introduces a knowledge elicitation framework to assess expert performance based on statistical accuracy and knowledge informativeness, highlighting substantial variations in experts’ abilities. Chapter 3 extends this analysis by exploring expert behaviour from a cognitive perspective, classifying forecasters into risk attitude groups (optimists vs. pessimists) and evaluating whether their predictions align with rational behaviour. Chapter 4 presents a hybrid framework combining machine learning (SVR, RF) and deep learning (DNN, LSTM) models to optimise the Bank of England’s external professional macroeconomic forecasts.<br/><br/>This approach offers an innovative solution to selecting optimal hyperparameters, a key challenge in machine learning, and demonstrates the effectiveness of these methods, even with limited data. The key contributions of this thesis lie in developing new methods to evaluate expert performance, including scoring forecasting accuracy and informativeness, introducing a cognitive perspective to forecasting behaviour, and advancing the application of machine learning in macroeconomic prediction. These findings enhance our understanding of expert biases, improve predictive accuracy, and offer practical implications for decision-making in economic forecasting.<br/>
Degree
thesis:*- Name dc:type.qualificationname
- Ph.D.
- Level dc:type.qualificationlevel
- doctoral
- Grantor dc:publisher.institution
- University of Southampton
- Year dc:date.issued
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Chang, Yujia
- Advisors dc:contributor.advisor
-
- Brito, Mario
- Mishra, Tapas
- Luo, Di
- ORCID person:identifier.orcid
Rights
- Language dc:language
- English, eng