Essays in empirical asset pricing

• Autor(a) principal: Baba Yara, Fahiz
• Data de Publicação: 2021
• Idioma: eng
• Título da fonte: Repositórios Científicos de Acesso Aberto de Portugal (RCAAP)
• Texto Completo: http://hdl.handle.net/10362/143131
• topic: Value Premium; Value Spread; Machine Learning; Neural Networks; Characteristic Sorted Portfolios; Cross-sectional Return Predictability; Domínio/Área Científica: Ciências Sociais: Economia e Gestão

The first chapter of this dissertation studies value strategies across equities, industries, commodities, currencies, global government bonds, and global stock indexes. We find that these strategies are predictable in the time series by the respective value spreads. A single component of the value spreads across asset classes capture about two-thirds of the value return predictability. The second chapter analyses returns to new and old sorts, where new (old) sorts capture the return of a characteristicsorted portfolio immediately (longer) after portfolio formation. We find that there exist large alphas between old and new sorts. These alphas (i) translate into large improvements in Sharpe ratio, (ii) are not captured by benchmark asset pricing models, and (iii) are linked to the return differential between new and old stocks. The final chapter investigates how incorporating results from the financial-economics literature in the specification of a machine learning model can improve the resulting models’ forecasts. I find that the economically motivated specification predicts better cross-sectional variation in individual stock returns and more robustly predicts time-series variation in returns to value-weighted long-short portfolios and the market portfolio.