Quantpedia Premium Update – 5th February

New strategies:

#825 – Hedging Factor in Cryptocurrencies

Period of rebalancing: Weekly
Markets traded: cryptos
Instruments used for trading: cryptos, ETFs, futures
Complexity: Moderately complex strategy
Backtest period: 2019-2022
Indicative performance: 4.68%
Estimated volatility: not known

Source paper:

Dunbar, Kwamie and Owusu-Amoako, Johnson: Predictability of Crypto Returns: A Habit-Based Explanation of the Risk Premium
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4288808
Abstract:
We evaluated the ability of futures market participants’ hedging decisions to predict changes in cryptocurrency returns based on its influence on risk aversion via the risk premium channel. We document that the hedging factor has a significant effect on measures of risk aversion and financial market uncertainty. Notably, the hedging factor displayed a more significant effect on risk aversion relative to measures of uncertainty. The out-of-sample evidence also suggests significant return predictability by the hedging factor. Further, our findings indicate that the hedging factor offers significant but modest gains during “normal” periods and significant but outstanding gains in periods characterized by acute macroeconomic stress.

#826 – News Sentiment and Equity Returns – BERT ML Model

Period of rebalancing: Daily
Markets traded: equities
Instruments used for trading: stocks
Complexity: Very complex strategy
Backtest period: 2002 – 2020
Indicative performance: 1.29%
Estimated volatility: 41.35%

Source paper:

Dangl, Thomas and Salbrechter, Stefan: News Sentiment and Equity Returns
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3971880
Abstract:
We investigate the impact of financial news on equity returns and introduce a non-parametric model to generate a sentiment signal, which is then used as a predictor for short-term, single-stock equity return forecasts. We build on Google’s BERT model (for Bidirectional Encoder Representations for Transform ers, see Devlin et al., 2018) and sequentially train it on financial news from Thom son Reuters1 covering the period from 1996 to 2020. With daily return data of S&P 500 constituents, our analysis shows that financial news carry information that is not immediately reflected in equity prices. News is largely priced-in within one day, with diffusion varying across industries. We test a simple trading strat egy based on the sentiment signal and report a return per trade of 24.06 bps and significant alpha of 77.56% p.a. with respect to a Fama-French 5-factor model plus momentum over an 18 year out-of-sample period.

#827 – A Machine Learning Approach to Stock Returns Prediction in China

Period of rebalancing: Monthly
Markets traded: equities
Instruments used for trading: stocks
Complexity: Very complex strategy
Backtest period: 1997 – 2019
Indicative performance: 26.68%
Estimated volatility: 23.88%

Source paper:

Huihang Wu, Xingkong Wei, Xiaoyan Zhang: Are Stock Returns Predictable in China? A Machine Learning Approach
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3971419
Abstract:
The predictability of stock returns has always been one of the core research questions in finance. This paper attempts to introduce machine learning method to answer whether stock returns are predictable in China. With 108 characteristics data in Chinese stock market from January 1997 to December 2019, this paper compares the out of sample predictability of the traditional econometric model with that of 6 major machine learning models, including partial least squares, principal component regression, elastic net regression, random forests, gradient boosted regression trees and neural networks. The main findings of this study are as follows: (1) historical trading data can predict individual stock returns in the next month, and the out-of-sample prediction of machine learning algorithm is better than that of traditional econometrics model; (2) in Chinese stock market, liquidity characteristics have strong predictive power, while momentum characteristics are weak in out of sample prediction; (3) the combination of machine learning algorithm and asset pricing research can generate significant economic value. During the out of sample test period, the performance of two layer neural network equal-weighted (value-weighted) long-short strategy is the best among all models, with an average monthly return of 3.03% (2.94%), the monthly volatility of 4.65% (6.88%), the annualized Sharpe ratio of 2.26 (1.48), and the significant monthly adjusted Alpha of 3.03 (2.95) in terms of FF5 factor. We present results that demonstrate that machine learning algorithm have clear merit over traditional techniques.

#828 – Time-Series Momentum Portfolios with Deep Multi-Task Learning

Period of rebalancing: Monthly
Markets traded: bonds, commodities, currencies, equities
Instruments used for trading: CFDs, futures
Complexity: Very complex strategy
Backtest period: 2000 – 2020
Indicative performance: 7.9%
Estimated volatility: 9.75%

Source paper:

Ong, Joel and Herremans, Dorien: Constructing Time-Series Momentum Portfolios with Deep Multi-Task Learning
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4288770
Abstract
A diversified risk-adjusted time-series momentum (TSMOM) portfolio can deliver substantial abnormal returns and offer some degree of tail risk protection during extreme market events. The performance of existing TSMOM strategies, however, relies not only on the quality of the momentum signal but also on the efficacy of the volatility estimator. Yet many of the existing studies have always considered these two factors to be independent. Inspired by recent progress in Multi-Task Learning (MTL), we present a new approach using MTL in a deep neural network architecture that jointly learns portfolio construction and various auxiliary tasks, such as forecasting realized volatility. Through backtesting from January 2000 to December 2020 on a diversified portfolio of continuous futures contracts, we demonstrate that even after accounting for transaction costs of up to 3 basis points, our approach outperforms existing TSMOM strategies. Moreover, experiments confirm that adding auxiliary tasks indeed boosts the portfolio’s performance. These findings demonstrate that MTL can be a powerful tool in finance.

#829 – Environmental Machine Learning Strategy in Equities

Period of rebalancing: Monthly
Markets traded: equities
Instruments used for trading: stocks
Complexity: Very complex strategy
Backtest period: 2011 – 2020
Indicative performance: 16.08%
Estimated volatility: 12.86%

Source paper:

Brown, William O. and Gao, Xiaoli and Han, Yufeng and Huang, Dayong and Wang, Fang: Environmental Variables and Stock Returns
https://ssrn.com/abstract=4232425
Abstract
Individual environmental variables may contain information that is obscured in aggregate environmental scores. We apply machine learning methods to granular environmental variables and study whether they are associated with future stock returns in the cross-section. A long-short portfolio that longs stocks with high forecasted returns and sells stocks with low forecasted returns earns more than one percent per month. Stocks with high forecasted returns are associated with high environmental operational standards. One interesting finding is that Scope 3 emissions are more important than Scope 1 and Scope 2 emissions in predicting future stock returns. Consistent with Pastor, Stambaugh, and Taylor (2022), the long-short portfolio performs better when climate concerns in the media are more intense.

New research papers related to existing strategies:

#649 – Scaled Volume in Cryptos

Babiak, Mykola and Erdis, Mustafa Berke: Variations in Trading Activity, Costly Arbitrage, and Cryptocurrency Returns
https://ssrn.com/abstract=4291073
Abstract:
Motivated by the cross-sectional and temporal fluctuations in their trading activity, digital assets provide a fruitful environment to investigate the impact of trading activity on expected returns. Portfolio-level analyses and cryptocurrency-level cross- sectional regressions demonstrate a negative and significant relation between expected returns and variability of dollar trading volume and turnover. This effect is robust to controls for beta, size, momentum, network premium, lottery demand, liquidity, idiosyncratic, volatility, and downside risks. The effect of trading activity variation is more pronounced among cryptocurrencies that are costlier to arbitrage. A zero-cost strategy exploiting this predictability survives transaction costs, various size and liquidity screens.

#636 – Machine Learning and Stock Anomalies in China

Wang, Jianqiu and Wang, Zhuo and Wu, Ke: Forecasting Stock Market Return with Anomalies: Evidence from China
https://ssrn.com/abstract=4282008
Abstract:
We empirically investigate the relation between anomaly portfolio returns and market return predictability in the Chinese stock market. Using 132 long-leg, short-leg, and long-short anomaly portfolio returns, we employ several shrinkage-based statistical learning methods to capture predictive signals of the anomalies in a high-dimensional setting. We find statistically and economically significant return predictability using long- and short-leg anomaly portfolio returns. Moreover, high arbitrage risk enhances forecasting performance, supporting that the predictability stems from mispricing correction persistence. Unlike the U.S. stock market, we find little evidence that the long-short anomaly portfolios can help predict market return due to the low persistence of asymmetric mispricing correction. We provide simulation evidence to sharpen our understanding of the differences found in the U.S. and Chinese stock markets.

Liu, Siyi and Xiang, Zhiqiang: Deep Learning Stock Portfolio Allocation in China: Treat Multi-Dimension Time-Series Data as Image
https://ssrn.com/abstract=4317790
Abstract:
A deep learning method is applied to predict stock portfolio allocation in the Chinese stock market. We use 6 original price and volume series as benchmark model settings and further explore the model’s predictive performance with social media sentiment. Our results show that our model can achieve a high out-of-sample Sharp ratio and annual return. Moreover, social media sentiment could increase the performance for both Sharp ratio and annual return while reducing annual volatility. We provide an end-to-end stock portfolio allocation model based on deep neural networks.

#799 – Machine Learning Volatility Targeting of Equity Indices

Ryu, Doojin: Forecasting Stock Market Volatility and Application to Volatility Timing Portfolios
https://ssrn.com/abstract=4295692
Abstract:
This study predicts stock market volatility and applies them to the standard problem in finance, namely, asset allocation. Based on machine learning and model averaging approaches, we integrate the drivers’ predictive information to forecast market volatilities. Using various evaluation methods, we verify that those high-dimensional models have better predictive performance relative to the standard volatility models. Furthermore, we construct volatility timing portfolios and discover that portfolios based on high-dimensional models mostly yield higher Sharpe ratios compared with the market.

#160 – Long-Term PE Ratio Effect in Stocks
#207 – Value Factor – CAPE Effect within Countries

Hillenbrand, Sebastian and McCarthy, Odhrain: The Optimal Stock Valuation Ratio
https://ssrn.com/abstract=4288780
Abstract:
Stock valuation ratios contain expectations of returns, yet, their performance in predicting returns has been rather dismal. This is because of an omitted variable problem: valuation ratios also contain expectations of cash flow growth. Time-variation in cash flow volatility and a structural shift towards repurchases have magnified this omitted variable problem. We show theoretically and empirically that scaling prices by forward measures of cash flows can overcome this problem yielding optimal return predictors. We construct a new measure of the forward price-to-earnings ratio for the S&P index based on earnings forecasts using machine learning techniques. The out-of-sample explanatory power for predicting one-year aggregate returns with our forward price-to-earnings ratio ranges from 7% to 11%, thereby beating all other predictors and helping to resolve the out-of-sample predictability debate (Goyal and Welch, 2008).

And several interesting free blog posts have been published during last 2 weeks:

Size Factor vs. Monetary Policy Regime

We have brought attention to the importance of evaluating factors models in different market regimes, and now, we will take a closer look at the size factor. Size [SMB (small minus big)] factor is a popular investment choice for asset investigation by many portfolio managers worldwide. The Size earned prominence in Fama and French’s three and five-factor models, and enjoy the continued discussion about its place in today’s portfolio construction. But it’s crucially important for investors seeking to capture the Size premium to realize that it is dependent on the monetary policy being pursued by the Federal Reserve, as the monetary easing seems to induce a Size premium.

160 Years of Wars and Disasters in Markets

Life is not always rosy; many tragedies and unexpected events hurt individuals and society. While some are hardly avoidable, such as natural disasters, some others as wars, are generally only functions of hate and greed. In the case of predictable events, risk measures can be employed, but unexpected outbreaks of aggression can hardly be hedged across the spectrum of different financial assets. We had previously touched on a similar topic and looked at some historical geopolitical shocks and price reactions around that time. Now, we would like to do a short review of an interesting 140-page paper by Dat Mai and Kuntara Pukthuanthong (2022), which, while not providing actionable strategy, provides insightful retrospection and takes war topic modeling to the higher level, covering developing narratives and influence factors extensively.

An Analysis of Rebalancing Performance Dispersion

The theme of rebalancing in longer-term investing is neglected but important as it influences the overall portfolio’s performance and risk. Unfortunately, many investors are inconsistent in choosing dates for their rebalances of portfolios, resulting in hardly predictable results (whether positively or negatively affecting it), and not contributing to handling risk management properly. The following article presents our analysis of the impact of rebalancing on portfolio returns. It also serves as an introduction to the methodology for an upcoming Quantpedia Pro report that our users would be able to use to quickly assess the impact of the rebalancing period on any selected combination of trading strategies, custom equity curves, and ETFs.

How to Use ETF Flows to Predict Subsequent Daily ETF Performance

Exchange-traded funds (ETFs) are incredibly versatile investment vehicles. They have become more popular in recent years as investors have grown more comfortable with passive investing strategies. But ETFs can be very useful also in active trading strategies, as they can be used to gain exposure to specific markets, sectors, or themes. But when you invest in ETFs or trade them regularly; it’s really good to look under the hood and learn some tricks where to obtain a new source of alpha. And one such possible source or information advantage may be the possibility of analyzing the ETF flows data …