Doctoral thesis: Cross-prediction can improve institutional investment decisions

Institutional investing is a constant balancing act between long-term performance and being affected by unfavourable market conditions. In his doctoral thesis, Studies on Cross-Prediction, Petteri Arinen examines how understanding the relationships between different assets’ returns over time can be used to make better predictions about future returns.

Although the co-movement of asset returns is central to classical financial theory, most foundational models rely on strongly simplifying assumptions. Arinen uses modern statistical, computational, and machine-learning techniques in his research.

– My analysis looks at how using cross-prediction to forecast returns works within different portfolio design and asset allocation models. Instead of relying on separate signals for each asset, this method combines shared information across assets, which consistently improves prediction accuracy, says Arinen.

The thesis comprises three complementary studies. According to Arinen, taken together they demonstrate that accounting for cross-sectional dependence among portfolio assets over time improves the accuracy of return predictions. These enhanced predictions improve the accuracy of both risk management and asset allocation and ultimately lead to better risk-adjusted investment returns.

You can read the whole thesis here: Studies on Cross-Prediction

Petteri Arinen will defend his doctoral thesis on 16 January at 12:00 at Hanken School of Economics, Arkadiankatu 22, Helsinki.
Opponent: Hossein Asgharian, Lund University
Chair: Anders Löflund, Hanken School of Economics

You can participate in the defence on-site.