Three Essays on Return Predictability and Asset Pricing
Abstract
Read the dissertation at https://ink.library.smu.edu.sg/etd_coll/232. Return predictability is important for tests of market efficiency and helps researchers to build better asset pricing models to explain the dynamics of asset prices. My dissertation contributes to the literature by analyzing the return predictability of technical indicators and investor sentiment, and investigate their implications for asset pricing and portfolio management. In Chapter 2, I study the predictive ability of a variety of technical indicators vis-á-vis the economic variables. I find that technical indicators have significant in both in- and out-of-sample forecasting power. Moreover, I find that using information from both technical indicators and economic variables increases the forecasting performance substantially. I also find that the economic value of bond risk premia forecasts from our methodology is comparable to that of equity risk premium forecasts. In Chapter 3, I find that market and size premiums are substantially higher following the up market than those following the down market, so that a portfolio could be mispriced by the unconditional asset pricing model, even if the conditional asset pricing models hold perfectly, if factor loadings vary over the up and down markets. I thus develop a trend-based conditional asset pricing framework, in which portfolios’ factor loadings are allowed to vary with the up and down markets. Empirically, I find that the trend-based conditional model largely explains the cross-section of technical analysis profitability anomaly in Han, Yang, and Zhou (2013), and the cross-sectional variation in technical analysis profitability appears to be driven by risk rather than mispricing. In Chapter 4, I propose a new sentiment index constructed with the purpose of predicting the aggregate stock market. In contrast with the widely used Baker and Wurgler (2006) sentiment index, our aligned index eliminates the common noise component of multiple sentiment proxies. Empirically, I find that the new index has greater power in predicting the aggregate stock market than the Baker and Wurgler (2006) index: it increases the predictive R2s by more than five times both in-sample and out-of-sample, and outperforms any of the well recognized macroeconomic variables. The predictability is both statistically and economically significant. Moreover, the new index improves substantially the forecasting power too for the cross-sectional stock returns formed on industry, size, value, and momentum. Finally, consistent with Baker and Wurgler (2007), I show that the driving force of the predictive power of investor sentiment stems from investors’ biased belief about future cash flows.