Dissecting the effectiveness of firm fundamentals in predicting cross-sectional stock returns

The stock market is one of the most important components of the capital market; therefore, it is crucial to ascertain what drives expected stock returns’ movements up and down, in order for the capital market to serve the real economy. Since the 1970s, many studies in asset pricing have suggested that differences in stock returns are due to differences in firm fundamentals, which are known as anomalies. Along with financial ratios increasingly being exploited, the mechanisms driving fundamental predictive power are becoming a hot topic in asset pricing. To interpret the large quantities of anomalies, the Fama and French (1993) an asset pricing model of three factors has been further developed into one of five factors (Fama and French, 2015) or four factors (Hou et al., 2017). Despite their different theoretical bases, both these two latter competing models consider profitability and investment as additional pricing factors, suggesting the great importance of these two groups of anomalies. Hence, the present study investigates the predictive power of firm investment in the stock market, and then goes further into how to integrate single profitability signals or large quantities of fundamental indicators into one proxy and tests their relation with expected stock returns. Moreover, this study tries to explain for the fundamental anomalies from both the micro and macro perspective. The research questions and results are as follows.

First, I examine the relations between firm investment and expected stock returns as well as the corresponding mechanisms driving these relations (Chapter 2). Based on the two-period model of the investment-q theory, the following research hypotheses are proposed: that firm investment is negatively correlated with expected stock returns, and that this negative relation is steepened among firms with greater investment friction. The empirical results show that company investment is indeed a negative predictor of stock returns in the US stock market, and that the greater the investment friction, the greater the premium generated by firm investment, which is especially significant in the investment-to-asset and asset-growth anomalies. It is also found that the equity constraint is likely to drive these two anomalies, and that the overall financing constraint also helps to explain the predictive power of investment-to-assets. Thus, the empirical results are consistent with the investment-q theory. For example, the firms, which have more difficulties issuing equities, usually face higher investment costs. Given the same increase in investment-to-assets, the corresponding expected stock return decreases more than for those having less difficulties in equity issuing, showing a stronger negative correlation between asset growth and expected return. Therefore, this study can support the investment-q theory empirically with text-based investment friction measurements to dissect firm investment anomalies. 

Meanwhile, this thesis also examines mispricing channels due to limits-to-arbitrage and finds that limited arbitrage helps us understand the investment anomalies of asset growth, investment growth, and net operating assets in the US stock market. Specifically, a higher level of limits-to-arbitrage is likely to be a signal of suffering from mispricing and to lead to a stronger negative relation between firm investment and expected stock returns. In fact, the economic mechanisms based on limits-to-arbitrage and the investment-q theory are not mutually exclusive but explain, respectively, anomalies from the points of view of investors and of firms. Additionally, results show that net stock issues and net operating assets perform significantly better following high sentiment, confirming market sentiment as an alternative explanation of investment anomalies. Therefore, this thesis finds that all of the aforementioned mechanisms contribute to the cross-sectional predictive power of firm investment in stock returns.

Second, this thesis studies Chinese stock return predictability resulting from firm financial strength as a representative profitability indicator and the economic mechanism for this predictability (Chapter 3). I construct a quarterly F-score to proxy firm financial strength as the sum of 9 signals based on single financial ratios measuring profitability, balance sheet structure, and operation efficiency from quarterly financial reports (including the first-quarterly, semi-annual, third-quarterly, and annual reports). Signals are equal to 1 if the firm’s ability is improved compared to the same quarter last year in terms of financial strength and 0 otherwise. The results of univariate portfolio analysis show that the expected stock returns are higher for firms with higher F-score and that the predictive power of F-score cannot be explained away by the Fama and French (2015) five factors, indicating that F-score information might not be contained in a single signal, for example, ROA (return on assets) or ROE (return on equity). This conjecture is verified by correlation analysis and independent bivariate portfolio sorting. After controlling for profitability and (/or) size, book-to-market ratio, and momentum, F-score is still significantly, positively associated with the expected returns, suggesting robust predictive power of firm financial strength. To explain for the premium generated by F-score, results show that SOEs (low investment friction) and high-turnover firms (high limits-to-arbitrage) can generate higher returns in the F-score long–short spread portfolios. Moreover, the premium is higher following high sentiment and low economic policy uncertainty (EPU). Therefore, both the rational investment-q theory and behavioral mispricing from the micro perspective, and economic policy uncertainty from the macro perspective help explain for the effectiveness of firm financial strength in the Chinese stock market.  

Third, this study constructs a comprehensive quality measure of fundamentals based on a large number of fundamental indicators, tests its predictive power in the cross-sectional expected stock return, and explores the mechanisms of this predictive power (Chapter 4). First, based on the Gordon Dividend Growth Model, a universe of firm characteristics consisting of 94 indicators is obtained by selecting the company's fundamental indicators from five aspects: profitability, growth ability, operational strategy, potential value, and security. I employ principal component analysis (PCA), Fama and MacBeth regression (FM), forecast combination (FC), scaled principal component analysis (SPCA), partial least squares (PLS) and the normalization approach (AFP) in Asness et al. (2019) to construct indicators to measure the comprehensive quality of firm fundamentals. The results of univariate portfolio sorting show that the predictor using PLS outperforms the others in predicting Chinese cross-sectional stock returns. In addition, after controlling for the value effect, investors can still earn money by longing the stocks of highest quality and shorting the counterparts. However, this investment strategy is only valid among small and micro stocks; that is, the less the market capitalization, the more profitable the strategy. This is doubly confirmed by the low and insignificant returns on the strategy in the large-cap universe.

For a better understanding of the predictive power of firm aggregated quality in stock returns, I examine the mispricing mechanisms based on investor sentiment and limits-to-arbitrage and the driving mechanisms of the business cycle and financial cycle. (i) Behavioral theory based on investor sentiment argues that the premium generated by firm quality should be higher among optimistic investors than their pessimistic counterparts. Empirical results show that the spread portfolio returns based on long and short strategies are higher following high than low market sentiment. After controlling for the market, size and value factors, the results are still robust, consistent with the theoretical expectations. (ii) Based on the mispricing channel of limited arbitrage, it is rational to conjecture that the predictive power of firm quality should be stronger among firms that find it difficult to arbitrage. The results of bivariate portfolio analysis show that the higher the level of limits-to-arbitrage (higher illiquidity, lower stock price, smaller trading volume), the larger the premium generated by fundamental quality. (iii) In terms of the macro cycles, results show that the quality premium is higher following recession periods, consistent with the risk compensation argument in Arshanapalli et al. (2006), and that no significant differences are found between rise and fall in financial markets. These findings indicate that the predictive power of firm quality is immune to systematic risk in financial markets but can, however, be explained by the macroeconomic cycle. In conclusion, the positive relation between the aggregated quality of firm fundamentals and the cross-sectional expected stock returns is driven by both mispricing at the micro level and business cycles at the macro level.