Return Predictability Research Articles

Retail Trading and Return Predictability in China

Abstract Using comprehensive account-level data, we separate Chinese retail investors into 5 groups and document strong heterogeneity in trading dynamics and performances. Retail investors with smaller account sizes cannot predict future returns correctly, display daily momentum patterns, fail to process public news, and show overconfidence and gambling preferences, while retail investors with larger account balances predict future returns correctly, display contrarian patterns, and incorporate public news in trading. Using performance measures established in previous literature, we find that smaller retail investors suffer from poor stock selection abilities and trading costs, while large retail investors’ stock selection abilities are offset by trading costs.

Do Consumption-Based Asset Pricing Models Explain the Dynamics of Stock Market Returns?

We show that three prominent consumption-based asset pricing models—the Bansal–Yaron, Campbell–Cochrane and Cecchetti–Lam–Mark models—cannot explain the dynamic properties of stock market returns. We show this by estimating these models with GMM, deriving ex-ante expected returns from them and then testing whether the difference between realised and expected returns is a martingale difference sequence, which it is not. Mincer–Zarnowitz regressions show that the models’ out-of-sample expected returns are systematically biased. Furthermore, semi-parametric tests of whether the models’ state variables are consistent with the degree of own-history predictability in stock returns suggest that only the Campbell–Cochrane habit variable may be able to explain return predictability, although the evidence on this is mixed.

The short-term predictability of returns in order book markets: A deep learning perspective

This paper uses deep learning techniques to conduct a systematic large-scale analysis of order book-driven predictability in high-frequency returns. First, we introduce a new and robust representation of the order book, the volume representation. Next, we conduct an extensive empirical experiment to address various questions regarding predictability. We investigate if and how far ahead there is predictability, the importance of a robust data representation, the advantages of multi-horizon modeling, and the presence of universal trading patterns. We use model confidence sets, which provide a formalized statistical inference framework well suited to answer these questions. Our findings show that at high frequencies, predictability in mid-price returns is not just present but ubiquitous. The performance of the deep learning models is strongly dependent on the choice of order book representation, and in this respect, the volume representation appears to have multiple practical advantages.

Noncommon Breaks

We develop a new Bayesian approach to estimate noncommon structural breaks in panel regression models. Any subset of the cross-section may be hit at different times within a break window. Break-specific parameters are learned from the cross-section. They reflect whether (i) breaks hit many or few series and (ii) there is a long or short lag between the first and final series hit by a break. In an empirical application to international stock return predictability, the method generates significantly more accurate forecasts than several benchmarks that yield economically meaningful utility gains for a risk averse investor with power utility.

Missing Data in Asset Pricing Panels

AbstractWe propose a simple and computationally attractive method to deal with missing data in in cross-sectional asset pricing using conditional mean imputations and weighted least squares, cast in a generalized method of moments (GMM) framework. This method allows us to use all observations with observed returns; it results in valid inference; and it can be applied in nonlinear and high-dimensional settings. In simulations, we find it performs almost as well as the efficient but computationally costly GMM estimator. We apply our procedure to a large panel of return predictors and find that it leads to improved out-of-sample predictability.

Persistence or reversal? The effects of abnormal trading volume on stock returns

Having established that portfolios derived from the extreme deciles of Abnormal Trading Volume (ATV) generate positive (negative) returns in the short (long) run, we devise a measure of Persistence in ATV (PATV) and provide an investor sentiment-based explanation for this return predictability. PATV leads to portfolio returns continuing to drift in the short run. However, the trading volumes of individual stocks in the portfolios gradually revert to their long-run means, accompanied by portfolio returns falling and turning negative as mispricing is corrected. We dismiss liquidity shocks, continuing overreaction, and investor disagreement and attention as explanations for the observed return predictability.

Common risk factors in cross-sectional FX options returns

Abstract We identify a comprehensive list of thirty-eight characteristics for predicting cross-sectional FX options returns. We find that three factors—long-term straddle momentum, implied volatility, and illiquidity—can generate economically and statistically significant risk premia not explained by other return predictors. Meanwhile, the predictability of the other characteristics becomes insignificant after accounting for the FX option three-factor model. The significance of the three factors is confirmed through a series of robustness tests covering different data sources, alternative options strategies, diversification effects, bootstrapping, and omitting crisis years.

EMPIRICAL TESTING OF FAMA-FRENCH ASSET PRICING MODEL IN INDONESIA STOCK EXCHANGE DURING COVID-19 PANDEMIC

The volatility of the Indonesian Stock Exchange (BEI) increased significantly during the Covid-19 pandemic period. In this period return predictability and price volatility in the stock index experienced a single structural break. There is concern among investors and academics that the asset pricing approach model that has been empirically accepted so far is unable to explain the return or excess return of an asset or investment during the Covid-19 pandemic period. This research tests the significance of the size (market capitalization), profitability, value (book-to-market), investment, and market risk premium (Rm-Rf) factors on the excess return of stock portfolios on the Indonesian Stock Exchange during the Covid-19 pandemic period. Existing studies show that the Covid-19 pandemic has affected investor sentiment, causing investors to panic and be pessimistic about their investments. In addition, there were deviations from the efficient market hypothesis during several pandemic periods in several countries so that stock prices did not fully reflect the available information. After testing, it was found that the factors size (market capitalization), profitability, value (book-to-market), investment, and market risk premium (Rm-Rf) did not have a significant influence on the excess return of stock portfolios on the Indonesia Stock Exchange during the period Covid-19 pandemic.

Testing the Forecasting Prowess of Bitcoin Uncertainty in the Predictability of Stock Returns

Motivated by the theoretical prediction of a weak link between the cryptocurrency market and stock markets, most empirical literature contends that Bitcoin is a safe hedge for the stock market. Opposing this position is the view that portrays Bitcoin as a regular, high-risk asset, such that when stock prices rise, so does Bitcoin, and vice versa. To test the validity or otherwise of this competing view, we construct a bivariate predictive model to examine the predictive power of Bitcoin uncertainty on stock returns. In contrast to the extant literature, we rely on the novel measure of uncertainty (i.e., the Bitcoin uncertainty indices, hereinafter URCY) and specify a forecasting model. The objectives of this study are to examine (i) the predictive power of UCRY indices on stock returns and (ii) the extent to which UCRY can make accurate out-of-sample forecasts. Using data for the G7 countries, among other things, our findings show that Bitcoin uncertainty indices are negative predictors of stock returns across the countries under investigation. Results also reveal that the indices are accurate and reliable predictors of stock returns in the short-to-medium term. These results are robust to accounting for structural breaks (i.e., the COVID-19 pandemic) and some macroeconomic variables. The policy implications of these results are discussed.

Picking a thorny rose: Optimal trading with spread‐based return predictability

AbstractSmall stocks' time‐varying spreads predict future return gap between small and large stocks. To optimally exploit such predictability, the investor captures current risk premium by purchasing at large spreads with substantially reduced turnover; uses an aim‐in‐front‐of‐the‐target approach to trade‐off between future risk premium and current transaction costs; and meets hedging demand at low costs. Strong interaction between transaction costs and return predictability leads to large losses from myopic trading. Greater variability of the spread is advantageous for investors who trade optimally but detrimental for investors who trade myopically. The spread‐based return predictability significantly increases the investment value of small stocks.

Geographic links and predictable returns

AbstractUsing establishment‐level data of U.S. public firms, we construct a novel measure of geographic linkage between firms. We show that the returns of geography‐linked firms have strong predictive power for focal firm returns and fundamentals. This effect is distinct from other cross‐firm return predictability and is not easily attributable to risk‐based explanations. It is more pronounced for focal firms that receive lower investor attention, are more costly to arbitrage, and during high sentiment periods. The cross‐firm information spillovers and return predictability are also stronger for geographic peers with economic linkages and with positive information. Our results are broadly consistent with sluggish price adjustment to nuanced geographic information.

Investor Regret and Stock Returns

We introduce a measure of regret for stock market investors and investigate its cross-sectional asset pricing implications. According to our regret-based framework, investors experience regret due to not achieving the highest possible return from a similar set of stock investments, and equity portfolios with high regret generate 6.84% more annualized alpha than portfolios with low regret. Using investor-trading activity of 78,000 households at a large U.S.-based brokerage firm, we develop an investor-based regret index and show that this household-level regret measure predicts stock returns in a similar way to our proposed regret measure. We also show that regret is not spanned by established risk or behavioral factors that have been documented to be robust predictors of equity returns. This paper was accepted by David Sraer, finance. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2022.03389 .

Forecasting stock prices changes using long-short term memory neural network with symbolic genetic programming

This study introduces an augmented Long-Short Term Memory (LSTM) neural network architecture, integrating Symbolic Genetic Programming (SGP), with the objective of forecasting cross-sectional price returns across a comprehensive dataset comprising 4500 listed stocks in the Chinese market over the period from 2014 to 2022. Using the S&P Alpha Pool Dataset for China as basic input, this architecture incorporates data augmentation and feature extraction techniques. The result of this study demonstrates significant improvements in Rank Information coefficient (Rank IC) and IC information ratio (ICIR) by 1128% and 5360% respectively when it is applied to fundamental indicators. For technical indicators, the hybrid model achieves a 206% increase in Rank IC and an impressive surge of 2752% in ICIR. Furthermore, the proposed hybrid SGP-LSTM model outperforms major Chinese stock indexes, generating average annualized excess returns of 31.00%, 24.48%, and 16.38% compared to the CSI 300 index, CSI 500 index, and the average portfolio, respectively. These findings highlight the effectiveness of SGP-LSTM model in improving the accuracy of cross-sectional stock return predictions and provide valuable insights for fund managers, traders, and financial analysts.

Stock Market Liquidity and Return Predictability: Investing in the long-run

Stock Market Liquidity and Return Predictability: Investing in the long-run

Large Language Models and Return Prediction in China

Large Language Models and Return Prediction in China

Happy App Happy Tip: The Return Predictability from Amusement Apps Downloads Around the World

Happy App Happy Tip: The Return Predictability from Amusement Apps Downloads Around the World

VIX constant maturity futures trading strategy: A walk-forward machine learning study.

This study employs seven advanced machine learning approaches to conduct numerical predictions of the next-day returns of VIX constant-maturity futures (VIX CMFs) using the term structure information derived from VIX CMFs. Based on precise numerical predictions, this study proposes a new Constrained-Mean-Variance Portfolio Optimization (C-MVO) trading strategy and tests it against a benchmark long-short trading strategy to evaluate the profitability of the machine learning numerical predictions. This study applies three unique feature sets, each incrementally incorporating the VIX CMFs' term structure features, to individually examine the predictive ability of the seven machine learning models and their backtesting performance. Over a comprehensive 11-year period, the experiment adheres to a strict walk-forward expanding-window methodology for both training and backtesting. The predictive and backtesting results show that four of the seven machine learning models attain a prediction information ratio greater than 0.02, with an average prediction information ratio of 0.037. This result suggests that the VIX CMFs term structure features have predictive power for the next-day returns of VIX CMFs. Moreover, the average C-MVO information ratio is 0.623, and the long-short strategy information ratio is 0.404. This increase in the information ratio under the C-MVO strategy validates the effectiveness of the machine learning models and the C-MVO strategy.

Enhancing Stock Return Predictions: Comparing Machine Learning Methods with Traditional Financial Models

This paper aims to understand if machine learning models can enhance stock price predictions compared to that of traditional financial models. The paper covers traditional financial models such as Stochastic Discount Factor models, factor-based models, option pricing models, and behavioural models, and machine learning techniques like supervised learning, deep learning, and hybrid models. By summarizing the results of various papers, this review compares the predictive accuracy of these models. The review found that machine learning methods, deep learning and hybrid models, outperformed traditional models as they captured nonlinear relationships between factors and stock price however found that some machine learning models were prone to overlearning. Through this review, financial analyst can understand if machine learning models should be used, which models to use specifically and can lead to enhanced stock price prediction.

Cryptocurrency return prediction: A machine learning analysis

Cryptocurrency return prediction: A machine learning analysis

Understanding rate of return dynamics of cryptocurrencies: an experimental campaign

In recent years, cryptocurrencies have been considered as an asset by public investors and received much research attention. It is a volatile asset, thus predicting its prices is not easy due to the dependence on multiple external factors. Machine learning models are becoming popular for cryptocurrency price predictions, while also considering social media data. In this article, we analyze the rate of return of three cryptocurrencies (Bitcoin, Ether, Binance) from an investor point of view. We also consider three traditional external variables: S&P 500 stock market index, gold price, and volatility index. The rate of return prediction is based on three stages. First, we analyze the correlation between the cryptocurrency returns and the traditional external variables. Next, we focus on the influential social media variables (from Twitter, Reddit, and Wikipedia). Later, we use these variables to improve prediction accuracy. Third, we test how the standard time series models (such as ARIMA and SARIMA) and four machine learning models (such as RNN, LSTM, GRU and Bi-LSTM) predict one-day rate of return. Finally, we also analyze the risk of investing in each cryptocurrencies using value risk statistics. Overall, our result shows no correlation between cryptocurrency returns and three traditional external variables. Second, we found that overall LSTM model is the best, GRU is the second-best prediction model, while the impact of the social media variables varies depending on the cryptocurrencies. Finally, we also found that investment in gold offers better returns than cryptocurrency during Covid-19-like situations.