Predict Market Returns Research Articles

Pooling and winsorizing machine learning forecasts to predict stock returns with high-dimensional data

We evaluate US market return predictability using a novel data set of several hundred ag- gregated firm-level characteristics. We apply LASSO, Elastic Net, Random Forest, Neural Net, Extreme Gradient Boosting, and Light Gradient Boosting Machine methods and find these models experience large prediction errors that lead to forecast failures. However, winsorizing and pooling machine learning model forecasts provides consistent out-of-sample predictability. To assess robustness, we apply machine learning methods to high-dimensional data for Canada, China, Germany and the UK as well as the Goyal–Welch data. All machine learning models we consider, except for the ensemble pooled methods, fail to significantly predict returns across our samples, highlighting the importance of pooling, evaluating additional economies, and the fragility of individual machine learning methods. Our results shed light on the sparsity versus density debate as the degree of sparsity and variable importance evolves over time.

Does the U.S. extreme indicator matter in stock markets? International evidence

We propose a new predictor—the innovation in the daily return minimum in the U.S. stock market (ΔMINUS\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\Delta {MIN}^{US}$$\\end{document})—for predicting international stock market returns. Using monthly data for a wide range of 17 MSCI international stock markets during the period spanning over half a century from January 1972 to July 2022, we find that ΔMINUS\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\Delta {MIN}^{US}$$\\end{document} have strong predictive power for returns in most international stock markets:ΔMINUS\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\Delta {MIN}^{US}$$\\end{document} negatively predicts the next-month stock market returns. The results remain robust after controlling for a number of macroeconomic predictors and conducting subsample and panel data analyses, indicating that ΔMINUS\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\Delta {MIN}^{US}$$\\end{document} has significant predictive power and it outperforms other variables in international markets. Notably, ΔMINUS\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\Delta {MIN}^{US}$$\\end{document} demonstrates excellent predictive power even during the periods driven by financial upheavals (e.g., Global Financial Crisis and European Sovereign Debt Crisis). Both panel regressions and out-of-sample tests also support the robust predictive performance of ΔMINUS\\documentclass[12pt]{minimal} \\usepackage{amsmath} \\usepackage{wasysym} \\usepackage{amsfonts} \\usepackage{amssymb} \\usepackage{amsbsy} \\usepackage{mathrsfs} \\usepackage{upgreek} \\setlength{\\oddsidemargin}{-69pt} \\begin{document}$$\\Delta {MIN}^{US}$$\\end{document}. The predictive power, however, disappears during the non-financial crisis caused by COVID-19 pandemic, which is originated from the health sector rather than the financial sector. The results provide a new perspective on U.S. extreme indicator in stock market return predictability.

Application of Machine Learning in Stock Market Return Forecasting

With the development of our country's financial market, people pay more attention to the stock market return forecast, which can bring the investors a lot of income and also become a tool of risk management. Machine learning (ML) and artificial intelligence (AI) are becoming increasingly sophisticated and have achieved amazing results in many fields. Machine learning and deep learning algorithm models can process huge amounts of data, from which they can quickly process and analyze laws, many scholars try to apply various machine learning models to the financial field. This paper first expounds the theoretical basis of stock market return prediction, and summarizes the main literature in this field in recent years, from the traditional stock prediction model to the machine learning prediction model. Finally, the traditional stock prediction model ARIMA model and the deep learning LSTM neural network model are selected for comparative empirical research. The sample data are from the Tushare big data community, taking the new energy vehicle stock BYD (002594) as an example. The empirical results show that the prediction accuracy and stability of LSTM neural network are far higher than ARIMA model, and it will have broad application prospects in financial prediction and other fields in the future.

Predicting consumption-wealth ratio changes and stock market returns

We show that the ability of Lettau and Ludvigson’s cay ratio to predict stock market returns could be greatly improved when adjustment is made for variations in expected cay ratio changes. We use the present-value identity of Campbell and Mankiw (1989) to derive the appropriate form of the adjustment and model the adjustment using a handful of macroeconomic variables. We find that the expected cay ratio changes vary in a predictable manner. Using our model to adjust cay improves the annual R̄2 values from 4.42% to 9.55% for in-sample prediction of excess market returns, and the out-of-sample pseudo R2 values increase from -1.88% and -1.72% to 2.38% and 3.69% with the first estimations using information up to 1985 and 1990, respectively. The superior performance also exists in medium-term market return predictions.

Predictability hidden by Anomalous Observations in Financial Data

Testing procedures for predictive regressions involving lagged autoregressive variables produce a suboptimal inference in presence of minor violations of ideal assumptions. A novel testing framework based on resampling methods that exhibits resistance to such violations and is reliable also in models with nearly integrated regressors is introduced. To achieve this objective, the robustness of resampling procedures for time series are defined by deriving new formulas quantifying their quantile breakdown point. For both the block bootstrap and subsampling, these formulas show a very low quantile breakdown point. To overcome this problem, a robust and fast resampling scheme applicable to a broad class of time series settings is proposed. This framework is also suitable for multi-predictor settings, particularly when the data only approximately conform to a predictive regression model. Monte Carlo simulations provide substantial evidence for the significant improvements offered by this robust approach. Using the proposed resampling methods, empirical coverages and rejection frequencies are very close to the nominal levels, both in the presence and absence of small deviations from the ideal model assumptions. Empirical analysis reveals robust evidence of market return predictability, previously obscured by anomalous observations, both in- and out-of-sample.

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

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

Climate Risk and Stock Markets: Implications for Market Efficiency and Return Predictability

ABSTRACT In this study, we set out to construct a daily frequency climate risk index with the aim to explore the impacts of climate risk on the stock markets of advanced and emerging countries and green stocks. We adopt an approach similar to the Sharpe ratio to demonstrate the pricing of climate risk into stocks and therefore reexamine the (in)efficiency of these markets. We specify an econometric model to evaluate the predictive content of the climate risk index for stock returns. We find compelling evidence of market efficiency when climate risk is priced into the stocks and market inefficiency otherwise. Results from the predictability analyses indicate that about 85% of advanced stock markets, 17% of emerging stock markets, and 100% of green stocks have the capacity to protect investors against climate risk. Forecast evaluations reveal that the climate risk index is a good predictor of the stock returns only after controlling for the macroeconomic environment. We explore several robustness checks and highlight both investment and policy implications.

The Virtue of Complexity in Return Prediction

ABSTRACTMuch of the extant literature predicts market returns with “simple” models that use only a few parameters. Contrary to conventional wisdom, we theoretically prove that simple models severely understate return predictability compared to “complex” models in which the number of parameters exceeds the number of observations. We empirically document the virtue of complexity in U.S. equity market return prediction. Our findings establish the rationale for modeling expected returns through machine learning.

Impact of the COVID-19 pandemic on the stock market and investor online word of mouth

Investor sentiment based on social media Word-of-Mouth (WoM) has been shown to be predictive of stock returns under normal market conditions. However, the COVID-19 pandemic has brought unprecedented impact and uncertainty to the stock market. In this paper, we examine the predictability of social media-based WoM on day-ahead US stock market returns across three phases of the pandemic: pre-COVID, peak-COVID, and recovery. We collect over 24 million stock-related tweets from Twitter across these three periods and employ text-mining methods for feature engineering. We focus on two channels of WoM information, aggregate sentiments, and disaggregated topics, representing investors' overall opinions and discussion patterns about the stock market. We find that, in contrast to normal market conditions, the investor sentiment indices are not significant in predicting market returns during the peak-COVID and recovery periods, while for the pre-COVID period, our findings are consistent with prior studies. To explore the effect of disaggregated latent topics, we propose an ensemble feature selection approach that identifies relevant topics for each of the three periods. We find that adding topic information significantly improves model prediction during the peak-COVID period. Our findings imply that during a global socio-economic disruption and the subsequent recovery, the landscape of financial markets, examined through the lens of social media-based WoM, is different than during normal market conditions. Our findings also suggest that investors who primarily trade on sentiment may wish to leverage machine learning techniques to identify latent topic information, subsequently incorporating such information into their trading decision-making.

Do Anomalies Really Predict Market Returns? New Data and New Evidence

Abstract Using new data from US and global markets, we revisit market risk premium predictability by equity anomalies. We apply a repertoire of machine-learning methods to forty-two countries to reach a simple conclusion: anomalies, as such, cannot predict aggregate market returns. Any ostensible evidence from the USA lacks external validity in two ways: it cannot be extended internationally and does not hold for alternative anomaly sets—regardless of the selection and design of factor strategies. The predictability—if any—originates from a handful of specific anomalies and depends heavily on seemingly minor methodological choices. Overall, our results challenge the view that anomalies as a group contain helpful information for forecasting market risk premia.

Islamic Fund Managers' Market Timing Abilities and their Impact on Fund Performances

This study examines Shariah-compliant fund's financial performance compared to conventional funds based on managerial timing abilities. Compared with traditional fund managers' timing skills, the study focuses on the most neglected area of Shariah-compliant manager timing abilities, liquidity, and volatility timing abilities. The study uses 69 funds whose significant exposure is the equity between January 2010 and June 2022, consisting of three investment styles: equity, asset allocation, and balanced funds. The study incorporates volatility factors based on Busse (1999) and liquidity factors from Amihud (2002) into the Carhart (1997) four-factor model. The study found that Islamic fund managers have better liquidity and volatility timing skills than conventional fund managers. The results are robust for analyzing individual funds based on different liquidity measures. There is no substantial evidence that Islamic and traditional fund managers can predict market returns.

Does the Adaptive Market Hypothesis Exist in Equity Market? Evidence from Pakistan Stock Exchange

In this study we empirically check the adaptive market hypothesis in Pakistan stock Market by measuring the association among present stock returns and past stock returns over the time of 2001 to 2020. We divided the weekly data of returns into subsamples of equal length 0f 3 years of seven samples and last sample with two years gap. For this purpose, we applied the five (linear and nonlinear) tests, in linear test Autocorrelation test, Variance ratio test, runs test and in nonlinear BDS independence test and Lagrange Multiplier test was applied to explicate that in what way the efficiency of market varies from time to time and whether their existence of any relationship between market condition and return predictability. Our results showed that efficiency of stock market fluctuatesamong episodic periods of dependency (inefficient condition) and independencies (efficient condition) in full and each subsamples thus it is conclude that PSX follow adaptive market and constant with AMH. Overall findings of the study concluded that AMH can better explain the stock return behavior then EMH. As the variation in market conditions can highly impress the trading activities and market efficiency so the investors can get the help from this study in making the investing decisions. It can be applied to practical setting such as asset allocation, investment consulting and risk management.

The optimized gate recurrent unit based on improved evolutionary algorithm to predict stock market returns

In order to accelerate the learning ability of neural network structure parameters and improve the prediction accuracy of deep learning algorithms, an evolutionary algorithm, based on a prior Gaussian mutation (PGM) operator, is proposed to optimize the structure parameters of a gated recurrent unit (GRU) neural network. In this algorithm, the sensitivity learning process of GRU model parameters into the Gaussian mutation operator, used the variance of the GRU model parameter training results as the Gaussian mutation variance to generate the optimal individual candidate set. Then, the optimal GRU neural network structure is constructed using the evolutionary algorithm of the prior Gaussian mutation operator. Moreover, the PGM-EA-GRU algorithm is applied to the prediction of stock market returns. Experiments show that the prediction model effectively overcomes the GRU neural network, quickly falling into a local optimum and slowly converging. Compared to the RF, SVR, RNN, LSTM, GRU, and EA-GRU benchmark models, the model significantly improves the searchability and prediction accuracy of the optimal network structure parameters. It also validates the effectiveness and the progressive nature of the PGM-EA-GRU model proposed in this paper with stock market return prediction.

Intraday Market Return Predictability Culled from the Factor Zoo

Intraday Market Return Predictability Culled from the Factor Zoo

Unintended look-ahead bias in out-of-sample forecasting

ABSTRACT This article shows that out-of-sample tests are susceptible to look-ahead bias not only to multiple testing problem that is emphasized in the literature. A forecaster often constructs a well-performing model without trial and error but with an intuition that is derived from observed empirical patterns in the test sample. Such an intuition, however, is unavailable in the beginning of the test sample. Therefore, the reported forecasting performance in an out-of-sample test is possibly exaggerated, although a forecaster simply utilizes her expertise without any intended p-hacking or fishing. A stylized forecasting model with an example of stock market return predictability quantitatively demonstrates this unintended look-ahead bias in out-of-sample tests.

A sentiment analysis approach to the prediction of market volatility.

Prediction and quantification of future volatility and returns play an important role in financial modeling, both in portfolio optimisation and risk management. Natural language processing today allows one to process news and social media comments to detect signals of investors' confidence. We have explored the relationship between sentiment extracted from financial news and tweets and FTSE100 movements. We investigated the strength of the correlation between sentiment measures on a given day and market volatility and returns observed the next day. We found that there is evidence of correlation between sentiment and stock market movements. Moreover, the sentiment captured from news headlines could be used as a signal to predict market returns; we also found that the same does not apply for volatility. However, for the sentiment found in Twitter comments we obtained, in a surprising finding, a correlation coefficient of -0.7 (p < 0.05), which indicates a strong negative correlation between negative sentiment captured from the tweets on a given day and the volatility observed the next day. It is important to keep in mind that stock volatility rises greatly when the market collapses but not symmetrically so when it goes up (the so-called leverage effect). We developed an accurate classifier for the prediction of market volatility in response to the arrival of new information by deploying topic modeling, based on Latent Dirichlet Allocation, in order to extract feature vectors from a collection of tweets and financial news. The obtained features were used as additional input to the classifier. Thanks to the combination of sentiment and topic modeling even on modest (essentially personal) architecture our classifier achieved a directional prediction accuracy for volatility of 63%.

Stock price movements: Evidence from global equity markets

This paper studies how market-level stock price movements respond to changes in expected future aggregate cashflows versus discount rates in individual global equity markets. We find that market-level dividend yields have strong forecasting power for aggregate dividend growth over a short term in most stock markets but their forecasting power becomes weak with investment horizon; their forecasting power for returns tends to exhibit an opposite pattern. The variance decomposition of dividend yields exhibits interesting cross-country heterogeneity. We further find that dividend smoothing and country-level disclosure and investor protection are two important drivers of the diverse predictability patterns across countries. • Short-horizon dividend growth predictability by aggregate dividend yields is generally strong. • Over long horizons aggregate dividend growth predictability goes weak. • Market return predictability by dividend yields exhibits an opposite pattern. • The variance decomposition of dividend yields exhibits cross-country heterogeneity. • Dividend smoothing and disclosure are two important drivers of diverse predictability patterns.

Belief-based momentum indicator and stock market return predictability

Li et al. (2022) propose a new momentum indicator that combines past returns and consistent belief information, and show that the indicator positively predicts cross-sectional stock returns. Based on the momentum indicator of Li et al. (2022), we further develop a conditional past return (CPR) indicator that additionally adds the direction information for the investors' consistent belief. We examine the effectiveness of CPR as a predictor for stock market returns. Our evidence shows that CPR significantly and positively predicts future one-month market returns. And CPR provides unique predictive information that is not related to the other popular predictors. The abundant out-of-sample evidence further supports CPR’s predictive ability. Additionally, we detect the asymmetric role of CPR in predicting market returns and find that much of the predictive ability of CPR is attributed to the interaction between the positive past returns and the positive consistent belief.

A closer look at the regime-switching evidence of bull and bear markets

Studies that use Markov-switching models to identify bull and bear markets indicate that conditional expected market returns display substantial variation over time. However, I find that in-sample tests fail to produce any evidence that these models predict market returns. To explain this finding, I show that the estimated expected return for the bull regime must differ from the estimated expected return for the bear regime in order to capture skewness. Because market returns display negative skewness, the Markov-switching estimates of conditional expected returns will typically vary over time, regardless of whether there is any actual variation in conditional expected returns.

Stock market return predictability: A combination forecast perspective

Based on traditional macroeconomic variables, this paper mainly investigates the predictability of these variables for stock market return. The empirical results show the mean combination forecast model can achieve superior out-of-sample performance than the other forecasting models for forecasting the stock market returns. In addition, the performances of the mean combination forecast model are also robust during different forecasting windows, different market conditions, and multi-step-ahead forecasts. Importantly, the mean combination forecast consistently generates higher CER gains than other models considering different investors' risk aversion coefficients and trading costs. This paper tries to provide more evidence of combination forecast model to predict stock market returns.