The realized information ratio and the cross-section of expected stock returns
Mehran Azimi (University of Massachusetts Boston)
This study examines the predictability of asset returns with the information ratio and its specific variant, the Sharpe ratio. We find that the realized Sharpe ratio (rsr) negatively predicts the cross-section of stock returns. Predictability is not due to the components of the rsr, i.e. past returns and volatility, nor to size, book-to-market, momentum and several variables related to the microstructure of the market and investor preferences. A short-term rsr-based factor portfolio generates alphas in the range of 0.67% to 0.89% per month compared to ten large factor models with t-stats well above 5. Our results suggest that A risk-return measure is a strong predictor of cross-sectional stock returns, stronger than risk and return. We provide evidence that the rsr is an indicator of stock-level sentiment. We find similar results using multiple information ratios and training periods of up to one year when controlling for momentum. The average monthly Carhart’s alpha of 60 factors constructed using five information ratios, each with training periods ranging from one to 12 months, is 0.63%. The average of the corresponding t-stats is 4.5. The results suggest that the predictive power of the information ratio, and of the rsr in particular, differs fundamentally from its constituents.
Economic fundamentals and stock market valuation: an approach based on the CAPE
Maria Ludovica Drudi and Federico Nucera (Bank of Italy)
This article estimates a fair value model, based on macroeconomic fundamentals, of Shiller’s cyclically-adjusted price-to-earnings ratio (CAPE). Performing a cross-country analysis, we find that CAPE – a widely used measure of stock market valuations – is, in general, positively related to economic growth and negatively related to the real long-term interest rate and measures of economic volatility calculated using industrial production and inflation data. Empirical evidence from predictive regressions of actual stock market returns indicates that deviations of CAPE from its estimated fair value are negatively related to future stock market returns. A prediction model based on these deviations outperforms, in many cases, a model based on CAPE levels both in-sample and out-of-sample.
War Speech and Cross Section of Expected Stock Returns
David A. Hirshleifer (Marshall School of Business, USC), et al.
A war-related factor model derived from textual analysis of media reports explains the cross-section of expected asset returns. Using a semi-supervised topic model to extract speech topics from 7,000,000 New York Times articles spanning 160 years, the War Factor predicts the cross section of test asset returns derived from traditional construction techniques and machine learning, and covering 138 anomalies. Our findings are consistent with assets that are good war risk hedges receiving lower risk premiums, or assets that are more positively sensitive to the prospect of war being more overvalued. The performance bonus on the war factor is added to the standard effects.
Do emotions influence investor behavior?
Ron Bird (University of Waikato), et al.
Despite much discussion in the psychology and marketing literature about how emotions influence decision-making, this area of analysis has been largely neglected in the financial economics literature. We address this significant discrepancy by using proxies for emotions drawn from news and social media to assess their influence on investment decisions and, ultimately, asset pricing. We find strong evidence to support that emotions influence investor decision-making and provide important insights into the nature of this relationship. In general, we find that these positive emotions such as confidence and optimism influence investors’ reactions more than negative emotions. Finally, emotions based on news media listings have a greater influence on stock market valuations than those based on social media listings.
Can ChatGPT predict stock price movements? Predictability of feedback and large language models
Alejandro Lopez-Lira and Yuehua Tang (University of Florida)
We examine the potential of ChatGPT and other major language models to predict stock market returns using headline sentiment analysis. We use ChatGPT to indicate whether a given headline is good, bad, or news irrelevant to company stock prices. We then calculate a numerical score and document a positive correlation between these “ChatGPT scores” and subsequent daily stock returns. Additionally, ChatGPT outperforms traditional sentiment analysis methods. We find that more basic models such as GPT-1, GPT-2, and BERT cannot accurately predict returns, indicating that return predictability is an emerging capability of complex models. ChatGPT-4’s implied Sharpe ratios are higher than those of ChatGPT-3; however, the latter model has larger total returns. Our results suggest that integrating advanced language models into the investment decision-making process can produce more accurate predictions and improve the performance of quantitative trading strategies. Predictability is focused on smaller stocks and larger on companies with bad news, consistent with arguments for limits to arbitrage rather than market inefficiencies.
Factor Momentum vs. Stock Price Momentum: A Review
Nusret Cakici (Fordham University), et al.
Is factor momentum driving stock price momentum? Inspired by recent findings in the United States, we revisit this relationship in 51 markets. The dynamic effect of factors remains strong, both within and between countries, regardless of the typical drivers of return predictability. However, its ability to capture stock momentum profits fundamentally depends on methodological choices and datasets. Therefore, factor momentum cannot robustly subsume equity momentum in global markets. On the contrary, the latter explains the former better than the reverse. Our findings challenge the view that momentum merely multiplies other factors rather than being a separate anomaly.
Risk premiums – The analysts’ point of view
Pascal Büsing and Hannes Mohrschladt (University of Münster)
We examine time and cross-sectional series of stock market risk premia from the perspective of financial analysts. Our new approach is based on the notion that analysts’ stock recommendations reflect both their subjective return expectations and their perception of stock risk. Thus, we can empirically derive the assumed risk premia from the recommendations and the expected returns implied by the target price. We show that the analysts’ assumed risk premia are strongly countercyclical such that their correlation with the VIX is 72%. Moreover, they predict future stock returns and are closely related to the price-dividend ratio and other cyclical state variables. Cross-sectionally, the assumed risk premia are comparatively large for high-beta, small, and value stocks, supporting a risk-based interpretation of these characteristics.
Learn how to use R for portfolio analysis
Quantitative analysis of the investment portfolio in R:
An introduction to R for modeling portfolio risk and return
By James Picerno