Date of Degree

2-2022

Document Type

Dissertation

Degree Name

Ph.D.

Program

Economics

Advisor

Chu-Ping C. Vijverberg

Committee Members

Wim Vijverberg

Christos Giannikos

Subject Categories

Econometrics | Economics | Finance

Keywords

Cross-section of stock returns, Empirical asset pricing, Anomalies, Spatial dependence, Principal components, Industry-rotation portfolio

Abstract

This thesis examines co-movement across industry return and value and momentum asset price anomalies through a new perspective and uses machine learning and spatial econometrics approaches. The first chapter examines the main approaches developed in the cross-section asset pricing literature for finding risk variables. The second chapter focuses on spatial co-movement across US industry returns. We show that spatial co-movement explains the variance in US industry returns after accounting for exposure to common variables, serial dynamics, and industry sector-specific characteristics using a dynamic spatial panel data model. The results show that an investment strategy that buys industry portfolios with high own-return and high spatially connected (neighboring) portfolio return and sells industry portfolios with low own-return and low spatially connected (neighboring) portfolio return generates an annual non-market return of approximately 8%. In the third chapter of the dissertation, we propose a multi-factor model in which the extra variables (apart from the standard market factor) are the innovation in each sparse principal component. Our findings demonstrate that our suggested hedging factors, which include Production (PR), Housing (H), Yield (Y), and Yield Spread (YS), explain a significant portion of the spread in average equity premia of the momentum portfolio deciles and value portfolio returns. In addition, the paper examines whether the multi-factor model is consistent with Merton's (1973) ICAPM framework.

Share

COinS