Essays on model averaging and forecasting

Date

2024-7

Author

Güneş, Hakan

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

166
views

15
downloads

Cite This

This thesis addresses the issue of model uncertainty in the context of financial asset forecasting. It comprises three papers. The first paper presents evidence of the in sample predictability while utilizing frequentist model averaging methods. Furthermore, we find evidence of out-of-sample forecast ability while utilizing direct and iterative one-month ahead approaches, using recursive and rolling window schemes with or without lagged dependent variables and for different out-of-samples. The second and the third papers forecast direct one-month ahead excess gold returns and utilize both univariate and multivariate models. While the second paper concentrates on the direction information obtained from the forecasts, the third paper focuses on the point forecasts of excess gold returns. The second paper demonstrates that an investor can benefit from the direction information obtained from the forecasts. Furthermore, the introduction of transaction costs while switching between assets suggests that it would be beneficial for the investor to consider them in the forecasting/decision step. The third paper employs a variety of methodologies and suggests combining shrinkage and the bagging approach to address the issue of model uncertainty. The bagging of adaptive Least Absolute Shrinkage and Selection Operator (𝑎𝑑𝑎𝐿𝐴𝑆𝑆𝑂) method outperforms other methods in terms of statistical and economic measures.

Subject Keywords

Forecasting, Model uncertainty, Frequentist model averaging, Shrinkage estimator, Bagging

URI

https://hdl.handle.net/11511/110099

Collections

Graduate School of Social Sciences, Thesis