Three Essays on Asset Pricing with Firm Characteristics

This thesis examines the relationship between expected stock returns and various types of firm characteristics. Chapter 1 decomposes the return on equity by DuPont analysis into five ratios: Tax Burden, Interest Burden, Margin, Turnover and Leverage. I utilise portfolio analysis and Fama-MacBeth regressions to evaluate the predictive performance on future stock returns. The results show that all popular asset pricing models fail to explain the long-short portfolio re-turns sorted by leverage. Interestingly, tax burden and turnover are the only two variables that can capture the expected stock returns in Fama- MacBeth regressions. Additionally, turnover is found to have a significant impact when including operating profitability, but not with other profitability proxies. Chapter 2 segments the proxies of value effects (book-to-market ratio, retained earnings-to-market ratio, and contributed capital-to-market ratio) into past five-year variables and compares the performance at five-year, one-year, and one-month lags. The results show that the retained earnings-to-market ratio is only significant with no lags, while the contributed capital growth is the most significant factor in the regressions with past information. Prior stock returns less than one year can have greater predictive power on future stock returns than the other. Additionally, updating the data monthly can have incremental improvements on the significance of price-scaled variables. Chapter 3 modifies Kelly et al. (2019) Instrumented Principal Component Analysis (IPCA) to a static version using a large-dimensional firm characteristic matrix to construct managed portfolio returns. I evaluate the relationship between characteristic-adjusted managed portfolios from IPCA and 126 macroeconomic variables using the three-pass method by Giglio and Xiu (2021). The outcomes reveal two significant macro factors associated with Price Index and the Market factor, whereas mimicking portfolio analysis shows that most of significant macroeconomic variables are related to Industry Production and Housing. The result also indicates that rank transformation with a 0.5 shift provides better estimation.