Forecast Stock Returns Research Articles

Excess Volatility in Professional Stock Return Forecasts

Excess Volatility in Professional Stock Return Forecasts

VOLATILITY FORECASTING AFTER REMOVING THE NOISE: DOES IT REALLY PAY OFF?

The analysis in this paper represents the first attempt to examine whether the forecasting performance of one of the GARCH models can be improved after removing the noise from the stock market returns. Doing the analysis on eight developed stock markets helps in providing reliable results that can be generalized to more stock markets. This paper then examines the ability of GARCH models to forecast stock return volatility under a range of statistical forecasts. We specifically employ the multiscale-based approach, namely the maximum overlap discrete wavelet transformation, to remove the noisy data from the return series. Based on related theoretical literature, we assume that noise trading makes the forecasting performance more misleading, especially at the short-term horizon. Our particular interest then is whether wavelet de-noising of the data before estimation affects the ability of the models to provide accurate forecasts at a short investment horizon. To de-noise the data, we use soft thresholding and Stein’s Unbiased Risk Estimator to obtain the decomposition level-based threshold limit. Our key results demonstrate that de-noising returns improves the accuracy of volatility forecasts regardless of whether we used statistical metrics or tests of equal predictive accuracy. Moreover, in terms of a particular volatility model, the asymmetric GARCH approach tends to be preferred although this result is not universal. Indeed, the central result from our analysis is that the process of de-noising is more important than the specific model.

Forecasting stock return volatility in data-rich environment: A new powerful predictor

We forecast stock return volatility by using the partial least squares approach that extract a powerful predictor from data-rich environment. Empirical results indicate that the new index has superior out-of-sample forecasting performance than the existing indexes, and the discovery is consistent with the in-sample predictive power. Specifically, the application of the new-index is extended to the allocation of investment portfolios to support mean–variance investors obtain considerable economic gains. In addition, our results are robust to various checks. Overall, our findings confirm that the partial least squares approach can effectively improve stock return volatility forecasts in a data-rich environment, successfully outperforming the competitive models and far surpassing the benchmark model.

Evaluating the performance of ensemble classifiers in stock returns prediction using effective features

Stock market prediction is considered as an important yet challenging aspect of financial analysis. The difficulty of forecasting arises from volatile and non-linear nature of stock market, which is affected by varied uncertain factors, ranging from financial ratios to macroeconomic indicators. Recent advances in machine learning, particularly ensembles, have made it possible for academic researchers and financial practitioners to forecast the stock market more efficiently. The novelty of this work is to evaluate how stock return in an oil-dependent country (i.e., Iran), which has been facing stagflation for a long time due to economic and political issues, is affected by fundamental and macroeconomic indicators. Our main objectives are to (1) find the most important fundamental and macroeconomic indicators that control the stock returns of companies listed on the Tehran Stock Exchange (TSE); (2) compare the performance of newly developed bagging- and boosting-based ensembles in predicting annual real stock returns of the TSE; and (3) develop multiclass classification models to forecast stock returns. Prior studies mainly focused on developing binary classification models, which simply predict whether stock returns will be positive or negative in the future. We, however, design multiclass classification models to provide more information for the investors and reduce the uncertainties associated with the prediction. To this end, we first provide a comprehensive list of 57 potential features affecting the stock returns. Next, the data are carefully preprocessed and fed to 14 different bagging- and boosting-based ensembles (e.g., Random Forest, LightGBM, XGBoost, Extra-Trees, AdaBoost, CatBoost) to predict the stock returns. The performance of ensembles is evaluated through different measures (e.g., accuracy, F-score, G-mean). We then propose a novel feature selection method to identify the most contributing features to the stock returns. Our proposed model identifies nearly 65% of 57 original features as redundancy, resulting in 20 most significant features. The selected features are fed to the mentioned ensembles to re-predict the stock returns. Finally, the performance of stock returns forecasts with and without selected features is compared. To design the ensembles, we employ the data from listed companies on the TSE for a 15-year period, spanning between 2005 and 2020. Results suggest that boosting ensembles, in general, outperform bagging-based methods. Among the boosting ensembles, XGBoost and AdaBoost provide the best and worst predictive performance, respectively. Among the bagging-like ensembles, Rotation Forest is the most accurate one, whereas Random Patches performs the worst. Further, our proposed feature selection approach effectively identifies the most representative features for stock returns prediction and can be used as a reliable framework for future investment decisions.

Ichimoku Cloud Forecasting Returns in the U.S.

Purpose: We show that the Ichimoku Cloud can forecast stock returns in the U.S., Canada, Germany, and U.K. Design/methodology/approach: We use a regression of next months index return regressed on the Ichimoku Cloud entry signal for price crossing above 9 periods, 26 period, 52 periods and a crossover between 9 and 26 periods. The regression slope coefficient is recorded as the risk premium return. We also record the t-statistic and R2 of the model. We note that T-statistics of 1.65 are statistically significant. R2 is economically significant with a value above .5 percent. Findings: This is showing real-time application how the current Ichimoku Cloud signal can predict tomorrow’s stock return. The strongest results occur for lagged values one period in the U.S. which shows initial justification to using the Ichimoku Cloud. We additionally show the Ichimoku Cloud entry signals are strong in regards to T-statistics and R2 when benchmarked on each of the equity markets in the U.S., Canada, Germany, and U.K. Research limitation/implications: The model only considers technical indicators for forecasting risk premium and could benefit from additional indicators or macro fundamentals. Originality/value: This is the first paper to use Ichimoku Cloud in the risk premium forecast framework.

Machine learning techniques for cross-sectional equity returns’ prediction

We compare the performance of the linear regression model, which is the current standard in science and practice for cross-sectional stock return forecasting, with that of machine learning methods, i.e., penalized linear models, support vector regression, random forests, gradient boosted trees and neural networks. Our analysis is based on monthly data on nearly 12,000 individual stocks from 16 European economies over almost 30 years from 1990 to 2019. We find that the prediction of stock returns can be decisively improved through machine learning methods. The outperformance of individual (combined) machine learning models over the benchmark model is approximately 0.6% (0.7%) per month for the full cross-section of stocks. Furthermore, we find no model breakdowns, which suggests that investors do not incur additional risk from using machine learning methods compared to the traditional benchmark approach. Additionally, the superior performance of machine learning models is not due to substantially higher portfolio turnover. Further analyses suggest that machine learning models generate their added value particularly in bear markets when the average investor tends to lose money. Our results indicate that future research and practice should make more intensive use of machine learning techniques with respect to stock return prediction.

Comparative response of global energy firm stocks to uncertainties from the crude oil market, stock market, and economic policy

In this study, a comparative analysis of the impacts of uncertainties induced by government economic policy and the crude oil and stock markets on the stock returns of 62 energy firms is carried out across different market states. We find that economic policy uncertainty (EPU) has a stronger influence than the two market-based uncertainties. However, the oil market uncertainty (OVX) outperforms the general stock market uncertainty (VIX) for the causality-in-mean, but the reverse holds for the causality-in-variance. Stronger causal impacts are also observed during the normal and bearish market states, in that order. Finally, more significant connection is attributed to the causality-in-mean than the causality-in-variance. Thus, uncertainties, especially the one induced by the economic policy, should not be jettisoned in modeling and forecasting stock returns. The regime-switching regression model also supports the role of uncertainties in predicting the stock returns of the energy firms, with particular emphasis on adverse impact in most regimes. The fact that causality is stronger at the middle and lower quantiles also suggests the poor risk-hedging power of the stocks at the normal and bearish market states.

Long-horizon stock valuation and return forecasts based on demographic projections

We incorporate low-frequency information from demographic variables into a simple predictive model to forecast stock valuations and returns using demographic projections. The demographics appear to be an important determinant of stock valuations, such as the dividend–price ratio. The availability of long-term demographic projections allows us to provide (very) long-horizon forecasts of stock market valuations and returns. We also exploit the strong contemporaneous correlation between returns and valuations to improve return forecasts — something which is not possible in a predictive regression with only lagged predictors. Extensive pseudo out-of-sample forecast comparisons and tests demonstrate the predictive value that an accurate demographic projection can deliver. Although the availability of historical Census Bureau projections is limited, we demonstrate that they could have been employed in real time to improve true long-horizon stock return prediction. We show how the model can be used to adjust predictions under alternative demographic assumptions, incorporating, for example, the demographic impact of COVID-19 or recent changes to immigration policy. • Demographic ratios help predict valuation ratios, such as the dividend–price ratio. • Models including future demographic projections can outperform a model containing only current demographic information. • Traditional predictive regression can be augmented to include predicted values of demographic ratios. • Demographic predictions improve both pseudo and true out-of-sample stock return prediction at the five-year horizon.

ANALYST FORECAST DISPERSION, INVESTOR SENTIMENT, AND THE CROSS-SECTION OF STOCK RETURNS

This paper explores the role investor sentiment plays in the relationship between analyst forecast dispersion and stock returns. With short sale constraints, stock prices are determined by the optimistic investors. There are more optimistic investors during the high sentiment periods when investors suffer more from psychological bias. This is the first paper to document that following the high sentiment periods, stocks with the most analyst forecast dispersion are overpriced, earning significantly negative returns, while those with the least analyst forecast dispersion are not overpriced as the degree of belief dispersion is low. However, following the low sentiment periods, both are not overpriced. A portfolio that longs the least dispersed stocks and shorts the most dispersed stocks yields significantly positive returns only following the high sentiment periods. My findings can potentially reconcile the puzzling risk effect and mispricing effect in the literature. The risk (mispricing) effect suggests a positive (negative) relation between analyst forecast dispersion and future stock returns. Presumably, the magnitude of the mispricing effect depends on the proportion of irrational investors and their bias, which is positively related to investor sentiment. During the high sentiment period, the mispricing effect takes over, and the overall effect is negative. During the low sentiment period, the percentage of irrational investors is mediate, and the mispricing effect and the risk effect counter each other, leading to insignificant relation.

Economic policy uncertainty and stock returns among OPEC members: evidence from feasible quasi-generalized least squares

We examine the predictive ability of economic policy uncertainty on stock returns of selected OPEC countries. In order to deal with certain statistical properties of the predictors, which include serial correlation, persistence, conditional heteroskedasticity, and endogeneity effects, wse utilize the Feasible Quasi-Generalized Least Squares (FQGLS) estimator in order to obtain accurate forecast estimates. As a precondition for forecast analysis, we conduct the predictability test, which shows that economic policy uncertainty is significant only for five countries, namely Kuwait, Nigeria, Saudi Arabia, United Arab Emirates, and Venezuela. Hence, we proceed with the main forecast analysis for only this set of countries. Our results are twofold. We first account for asymmetries in forecasting stock returns by comparing the forecast performance of the symmetric economic policy uncertainty-based predictive model with its asymmetric variant. On the other hand, we compare the performance of the best model from above with the standard ARFIMA model using an alternative forecast test. In both cases, we find that the asymmetric model yields the most accurate forecast returns for stock returns of the five countries. In essence, neglecting the role of asymmetries in forecasting stock returns can lead to bias results. Our findings are not only robust to different sample sizes (i.e., 50%, and 75%) and different forecast horizons (4, 8, and 12 months) but have important policy implications for policymakers and potential investors.

Predicting power of ticker search volume in Indian stock market

This study examines the ability of online ticker searches to serve as a valid proxy for investor sentiment and forecast stock returns and trading volumes in the Indian financial market. In contrast to the findings in prior research, we observe that ticker search volumes do not exhibit any predictive value for future excess stock returns. However, we find a weak but significant positive effect of ticker search volumes on trading volume with a two-week lag. A battery of robustness checks supports our findings. Our work warns the investors from possible misleading insights arising from search volume and stock returns related studies.

Machine learning portfolios with equal risk contributions: Evidence from the Brazilian market

We investigate the use of machine learning (ML) to forecast stock returns in the Brazilian market using a rich proprietary dataset. While ML portfolios can easily outperform the local market, the performance of long-short strategies using ML is hampered by the high volatility of the short portfolios. We show that an Equal Risk Contribution (ERC) approach significantly improves risk-adjusted returns. We further develop an ERC approach that combines multiple long-short strategies obtained with ML models, equalizing risk contributions across ML models, which outperforms, on a risk-adjusted basis, all individual ML long-short strategies, as well as alternative combinations of ML strategies.

Statistical Analysis Dow Jones Stock Index—Cumulative Return Gap and Finite Difference Method

This study was motivated by the poor performance of the current models used in stock return forecasting and aimed to improve the accuracy of the existing models in forecasting future stock returns. The current literature largely assumes that the residual term used in the existing model is white noise and, as such, has no valuable information. We exploit the valuable information contained in the residuals of the models in the context of cumulative return and construct a new cumulative return gap (CRG) model to overcome the weaknesses of the traditional cumulative abnormal returns (CAR) and buy-and-hold abnormal returns (BHAR) models. To deal with the residual items of the prediction model and improving the prediction accuracy, we also lead the finite difference (FD) method into the autoregressive (AR) model and autoregressive distributed lag (ARDL) model. The empirical results of the study show that the cumulative return (CR) model is better than the simple return model for stock return prediction. We found that the CRG model can improve prediction accuracy, the term of the residuals from the autoregressive analysis is very important in stock return prediction, and the FD model can improve prediction accuracy.

Practical Applications of The Best of Both Worlds: Forecasting US Equity Market Returns Using a Hybrid Machine Learning–Time Series Approach

In <b><i>The Best of Both Worlds: Forecasting US Equity Market Returns Using a Hybrid Machine Learning–Time Series Approach</i></b>, published in the Spring 2021 issue of <b><i>The Journal of Financial Data Science,</i></b><b>Haifeng Wang, Harshdeep Singh Ahluwalia, Roger Aliaga-Díaz, and Joseph Davis</b> (all at<b> Vanguard</b>) revisit the long-term equity return forecasting performance of a popular time-series model based on Shiller’s cyclically adjusted price-to-earnings (CAPE) ratio. Noting that its performance had deteriorated due to structural economic changes, they considered additional variables, econometric methods, and machine learning (ML) techniques. Using their domain knowledge in the data science arena, the authors sought to improve the out-of-sample predictive accuracy of long-run stock return forecasts. They compared the performance of various approaches, including linear regressions, vector autoregressive (VAR) models, random forests, gradient boosting machines, and gated recurrent units. They found that a two-step approach that estimates a VAR model with ML techniques and uses outputs in an additive component model of returns works best. Using an ensemble of ML techniques within a robust economic framework yields more accurate long-run return predictions than using any of the methods alone.

Forecasting stock return volatility: The role of shrinkage approaches in a data‐rich environment

AbstractThis paper employs the prevailing shrinkage approaches, the lasso, adaptive lasso, elastic net, and ridge regression to predict stock return volatility with a large set of variables. The out‐of‐sample results reveal that shrinkage approaches exhibit superior performance relative to the benchmark of the autoregressive model and a series of competing models in terms of the out‐of‐sample R‐square and the model confidence set. By using shrinkage methods to allocate portfolio, a mean–variance investor can obtain significant economic gains. Overall, our findings confirm that shrinkage approaches can effectively improve stock return volatility forecasting in a data‐rich environment.

Improving the predictability of stock returns with global financial cycle and oil price in oil-exporting African countries

In the face of frequent market turbulences leading to losses for investors, it becomes unarguably essential to determine factors capable of accurately forecasting stock returns. Therefore, this study considers predictors from the global arena by determining the predictive information substance of the global financial cycle (GFCy) and global oil price on the stock returns of African oil exporters. Summarily, our results show that the multi-factor predictive model, which jointly accounts for the GFCy indicator and oil price, offers better forecast performance than the baseline (historical average) and the single-factor predictive models, except in Kenya. Further looking into the importance of oil shocks, forecast performance obviously appears to be sensitive to different oil shocks, predominantly in Kenya and Nigeria. The multi-factor predictive models that include the different oil shocks perform better than the contending models in other countries, while mixed evidence is found for these two exceptional countries. Robustness is provided for the results through various checks, including alternative forecast tests, different sub-samples, and multiple forecast horizons.

The Effect of Quantitative Easing through Google Metrics on US Stock Indices

The purpose of this study is to investigate the fluctuations that occur in stock returns of US stock indices when there is an increase in the volume of Google internet searches for the phrase “quantitative easing” in the US. The exponential generalized autoregressive conditional heteroscedasticity model (EGARCH) was applied based on weekly data of stock indices using the three-factor model of Fama and French for the period of 1 January 2006 to 30 October 2020. The existence of a statistically significant relationship between searches and financial variables, especially in the stock market, is evident. The result is strong in three of the four stock indices studied. Specifically, the SVI index was statistically significant, with a positive trend for the S&amp;P 500 and Dow Jones indices and a negative trend for the VIX index. Investor focus on quantitative easing (QE), as determined by Google metrics, seems to calm stock market volatility and increase stock returns. Although there is a large body of research using Google Trends as a crowdsourcing method of forecasting stock returns, this paper is the first to examine the relationship between the increase in internet searches of “quantitative easing” and stock market returns.

Industry-level media tone and the cross-section of stock returns

This paper investigates the cross-sectional relation between industry-level media tone and expected stock returns in China. Using a machine learning technique to establish a proxy for industry-level media tone, we find that stocks in industries with more positive media tone earn significantly higher future returns than these with more negative media tone. Specifically, relative to stock-specific media tone, industry media tone plays a dominant role in forecasting stock returns. Moreover, the return premium for higher industry-level media tone continues for two months, and the returns are positive though insignificant within half a year, indicating that the media contains fundamental information.

Good volatility, bad volatility, and time series return predictability

We propose a least squares estimator weighted by a combination of lagged realized semivariances related to positive and negative returns (WLS-CRS) and use univariate models alone and in combination to reveal significant return predictability. For an investor with a mean-variance preference who allocates a portfolio based on an equal-weighted combination of WLS-CRS model forecasts, the annual certainty equivalent return is 242.8 basis points higher than that received by an investor whose portfolio is allocated based on historical average forecasts. In forecasting stock returns, WLS-CRS estimates outperform the popular ordinary least squares estimates in both statistical and economic evaluation frameworks. WLS-CRS also outperforms estimators based on least squares weighted by lagged realized volatility. We further demonstrate the dominant role of negative return semivariance in improved forecasting performance. Our main findings hold through several robustness checks, including alternative validation samples, different risk aversion coefficients, and various forecast combinations.

Estimating Stock Returns Using Rough Set Theory: An Exploratory study With An Evidence From Iraq Stock Exchange

‎ This research aims to estimate stock returns, according to the ‎Rough Set Theory ‎approach, ‎test ‎its effectiveness and accuracy in predicting stock returns and their potential in the ‎field of ‎financial ‎markets, and rationalize investor decisions. The research sample is totaling (10) ‎companies traded at Iraq Stock Exchange. The results showed a remarkable ‎ ‎Rough Set Theory application in data reduction, contributing to the rationalization of ‎investment ‎decisions. The most prominent conclusions are the capability of rough set theory ‎in ‎dealing with financial data and applying it for forecasting stock ‎returns.‎The ‎research provides those interested in investing stocks in financial markets ‎with ‎significant financial analysis tools that exceed the traditional statistical methods. ‎The ‎originality ‎of the research lies in the diversification of financial and statistical analysis tools ‎and ‎methods of ‎forecasting stock returns