Two essays in stock return predictability
Loading...
Authors
Meeting name
Sponsors
Date
Journal Title
Format
Thesis
Subject
Abstract
This dissertation consists of two essays on the stock return predictability. The first essay in Chapter 1 titled Machine Learning Classification and Portfolio Allocation examines machine learning predictability using classification models and investigates the relation between predictability and stock returns. The classifiers show significant and robust out-of-sample precision in placing stocks into the correct deciles, outperforming their counterpart machine learning regressions. The corresponding long-short portfolios deliver significant [alpha] economic gains. The classifiers invest more resources in return state transitions with lower information shortage--and excel in predicting return deciles in the center and edges of the transition probability matrix. The classifiers extract information from different firm characteristics. Following Easley and O'Hara (2004), I show that prediction success is negatively related to the future returns at the stock level, controlling for information shortage. Information shortage also reduces the probability of prediction success. Portfolios conditional on information shortage show enhanced performance. A mimicking portfolio based on the shock of prediction precision generates significant benchmarking against popular factor models. The second essay in Chapter 2 titled 150 Years of Return Predictability Around the World: A Holistic View Across Assets studies the time series predictability in bond, stock, and real estate market using payout-price ratios such as dividend-price ratio for stock return forecast. Campbell and Shiller (1988a, b) show that if payout growth is not predictable, the payout-price ratio decides returns and the returns must be predictable. Using 150-year data from 16 developed countries across bond, equity, and housing markets, I revisit this implication using the payout-price ratios and study the payout-price return predictability across assets and countries. None of the 48 country-asset combinations shows consistent in-sample and out-of-sample performance with positive utility gain for the mean-variance investor. Contrary to Cochrane's finding, the VAR simulation using data from all the countries in the past 150 years does not reject the null that the dividend growth is predictable, while 14 (5) countries have in-sample predictable payout growth in the equity (housing) markets. Overall, the joint hypothesis test provides weak support to return predictability based on payout-price ratios.
Table of Contents
PubMed ID
Degree
Ph. D.