Promotie H. Zhang

Many of the sophisticated models for stock return forecasting and portfolio optimisation cannot beat naive equal-weighted models. The challenge is that, even in the age of big data, there are usually more potential variables than is appropriate for estimation. This thesis is dedicated to improving asset pricing models via ensemble learning methods without requiring more data.

By introducing a simple ensemble method, first, several representative sophisticated models of stock return forecasting are compared based on standard economic variables in the literature. The results demonstrated that, although these models, including standard machine learning methods such as LASSO and elastic net, underperform the equal-weighted combination of individual predictors, their performance improves dramatically when estimated under the new simple ensemble scheme, taking parameter instability and shrinkage explicitly into account.

Then, a more sophisticated ensemble method that averages forecasts from sophisticated models was introduced based on random subsamples and learning from prediction errors by adaptively changing sampling distributions. The results demonstrated that this method could significantly improve these sophisticated models. Its forecast gains stem from better performance during periods of market uncertainty and crises, and increased diversity and built-in shrinkage.

Finally, a general boosting framework for high-dimensional portfolio optimisation, where the classical mean-variance portfolios cannot work properly, was introduced. The results demonstrated the effectiveness of this study’s boosting methods in both low- and high-dimensional settings and found that they can outperform the 1/N portfolio in terms of several popular metrics.

• Locatie: Cobbenhagen building, Portrettenzaal
• Promotores: prof. dr. B. Jacobsen, prof. dr. F. Jiang, prof. Q. He