Quantpedia Premium Update – August 21st

New strategies:

#1162 – Relief Factor in the Cross-Section of Stocks

Period of rebalancing: Monthly
Markets traded: equities
Instruments used for trading: stocks
Complexity: Complex strategy
Backtest period: 1963-2022
Indicative performance: 10.39%
Estimated volatility: 13.66%

Source paper:

Gempesaw, David and Goldie, Brad and Henry, Tyler and Kassa, Haimanot: Relative Loss Aversion and the Cross-Section of Stock Returns
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5215160
Abstract: Drawing on relative income utility and loss aversion from prospect theory, we investigate whether poor-performing peer stocks shape investor preferences. Defining “relief” as the difference between a stock’s return and the worst-performing stock in its industry, we find that high-relief stocks have lower future returns than low-relief stocks. This effect strengthens with low relief values, weaker industry performance, and greater characteristic similarity to the benchmark. Relief exhibits limited long-term predictive power and lacks persistence, suggesting the benchmark’s relevance is short-lived. Our findings illustrate the role of positive relative performance comparisons in determining expected returns.

#1163 – Cultural Calendars and the Gold Drift: Are Holidays Moving GLD ETF?

Period of rebalancing: Daily
Markets traded: commodities
Instruments used for trading: ETFs, futures, CFDs
Complexity: Simple strategy
Backtest period: 2005-2025
Indicative performance: 2.34%
Estimated volatility: 3.31%

Source paper:

Vojtko, Radovan and Dujava, Cyril: Cultural Calendars and the Gold Drift: Are Holidays Moving GLD ETF
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5384405
Abstract: Financial markets exhibit persistent calendar anomalies, which often defy the efficient-market hypothesis by generating predictable return patterns tied to institutional or cultural events. In this paper, we document a novel, globally pervasive drift in gold prices surrounding major wealth-oriented festivals across the four principal cultural and religious domains: Christianity, Islam, Hinduism, and East Asian syncretic traditions. While each community endows its principal holidays with gift-giving rituals and conspicuous displays of wealth, the sole differentiator among regions is the precise timing of these festivities on the Gregorian calendar. Our central thesis is that gold, owing to its dual role as a universal wealth reservoir and socio-cultural status symbol, experiences concentrated, holiday-induced buying pressure that yields persistent and economically material drift in the GLD ETF. By quantifying this effect across four distinct cultural calendars, we introduce a previously undocumented demand-side factor into commodity-pricing models.

#1164 – Statistical Arbitrage in Rank Space

Period of rebalancing: Intraday
Markets traded: equities
Instruments used for trading: stocks
Complexity: Very complex strategy
Backtest period: 2006-2022
Indicative performance: 35.68%
Estimated volatility: 10.87%

Source paper:

Li, Y.-F and Papanicolaou, G.: Statistical Arbitrage in Rank Space
https://arxiv.org/abs/2410.06568
Abstract: Equity market dynamics are conventionally investigated in name space where stocks are indexed by company names. In contrast, by indexing stocks based on their ranks in capitalization, we gain a different perspective of market dynamics in rank space. Here, we demonstrate the superior performance of statistical arbitrage in rank space over name space, driven by a robust market representation and enhanced mean-reverting properties of residual returns in rank space. Our statistical arbitrage algorithm features an intraday rebalancing mechanism for effective conversion between portfolios in name and rank space. We explore statistical arbitrage with and without neural networks in both name and rank space and show that the portfolios obtained in rank space with neural networks significantly outperform those in name space.

#1165 – Statistical Arbitrage via Multi-View Spectral Clustering on Mixed Frequency Data

Period of rebalancing: Daily
Markets traded: equities, currencies, commodities, bonds
Instruments used for trading: stocks, futures, ETFs
Complexity: Very complex strategy
Backtest period: 2000-2022
Indicative performance: 6.8%
Estimated volatility: 8.19%

Source paper:

Leung, Raymond C. W.: Statistical Arbitrage via Single-view and Multi-view Spectral Clustering on Mixed Frequency Data
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4975855
Abstract: Systematically identifying clusters of similar assets is a critical step in statistical arbitrage strategies. This paper makes two key contributions: (i) In addition to assets’ daily closing prices, we incorporate realized estimators (i.e. realized volatility, realized beta, and others) derived from high-frequency intraday data into the daily feature set; and (ii) We examine the effectiveness of single-view and multi-view spectral clustering algorithms on these feature sets for identifying similar assets. We use these clustered assets to construct a variety of signals and trading rules, including the distance method, cointegration method, copula method, and an optimal entry-exit rule based on Ornstein-Uhlenbeck spread dynamics. We evaluate our methodology on the S&P 500 equities, futures contracts, exchange traded funds, and foreign exchange asset classes. Our findings show that trading profitability is influenced more by the selection of feature sets and clustering methods than by the choice of signals or trading rules. Specifically, strategies employing multi-view clustering algorithms, which integrate various realized estimators, consistently deliver superior risk-adjusted returns and reduced drawdowns compared to those based on the distance method and single-view clustering that rely solely on daily closing price data.

#1166 – Genetic Mimicking Portfolios for ETF Arbitrage

Period of rebalancing: Daily
Markets traded: bonds
Instruments used for trading: ETFs
Complexity: Very complex strategy
Backtest period: 2010-2021
Indicative performance: 29.12%
Estimated volatility: 9.4%

Source paper:

Crego, Julio A. and Kvaerner, Jens Soerlie and Sommervoll, Aavald and Sommervoll, Dag Einar and Stevens, Niek: Genetic Mimicking Portfolios for ETF Arbitrage
https://ssrn.com/abstract=5302810
Abstract: Financial markets are full of securities whose price depends on other securities that are often diﬃcult to trade. We develop a method to identify a liquid mimicking portfolio that tracks a de facto non-tradable asset. Our method combines a genetic algorithm with non-negative least squares. We apply it to the corporate ETF market. An arbitrage portfolio that takes a short position in ETF with the highest price relative to its NAV and a long position in the mimicking portfolio generates a Sharpe ratio of 3. The return on the mimicking portfolio correlates negatively with the return on the short position, is responsible for 1/3 of the Sharpe ratio, and reduces expected tail loss by 2/3.

New research papers related to existing strategies:

#626 – Image Recognition in Stock Price Charts Predicts Stock Returns

Chen, Zishan and Huang, Weige: Can Candlestick Charts Help Predict Stock Price Movement? A Machine Learning Approach
https://ssrn.com/abstract=5000007
Abstract: We utilize a convolutional neural network (CNN) to train a dataset of candlestick charts, aiming to predict the movement of the index prices in both the Chinese and US stock markets. We then compare the results and find that candlestick charts in the Chinese stock market contain more information for predicting the stock index movement. Furthermore, we discover that the prediction accuracy in the Chinese stock market is higher during periods of historical market success. However, we do not observe these trends in the US stock market. Since CNN can emulate the process of investors observing candlestick charts to make investment decisions, we speculate that these differences are related to the structure of the investor base in the two countries’ stock markets. This could be attributed to the fact that candlestick charts have the potential to impact the perception and attention of individual investors through visual stimulation and emotional responses, thereby influencing the predictive efficacy of the model.

#447 – Logistic Regression and Momentum-Based Trading Strategy
#496 – Using Machine Learning to Pick the Right Combination of Risky and Risk-Free Asset
#581 – Equity Index and Risk-free Asset Allocation Using Neural Networks
#734 – Empirical Asset Pricing via Machine Learning

Hussain, Sabbor and Dramane, Thiombiano and Chen, Johui and Ali, Wajid and Abbas, Qamar: Stock Market Predictability: Ensemble Learning Triumphs Over Traditional Methods
https://ssrn.com/abstract=5290750
Abstract: This study explores the predictability of 30-day stock market returns using macroeconomic indicators, comparing statistical and machine learning models. We analyzed both regression models for continuous S&P 500 returns and classification for bullish or bearish market signals, The predictive performance of the models was evaluated using three metrics, accuracy, area under the ROC curve (AUC), and F-measure. The results show that machine learning models consistently outperform the classical regression model in predicting the direction of S&P 500 returns. The results show that ensemble methods, particularly random forests and bagging, consistently outperform traditional linear models like logistic regression and linear discriminant analysis (LDA). After feature selection and hyperparameter tuning, bagging surpassed in classification, while gradient boosting led in regression. The Diebold-Mariano (DM) test confirmed the statistical significance of these findings, reinforcing the superior predictive power of ensemble methods for stock market returns prediction.

#748 – Bear Beta Factor Investing Strategy

Doshi, Hitesh and Jacobs, Kris: A Cross-Sectional Decomposition of Firms’ Market Betas
https://ssrn.com/abstract=5279556
Abstract: We propose a new empirical framework for cross-sectional asset pricing. The framework generalizes and nests the CAPM. We decompose the firm’s traditional CAPM market beta into two components: a negative market beta, which contains the negative correlations between the return of the firm and other firms, and a positive market beta, which contains the positive correlations. The sum of the positive and negative betas is the total market beta, and we expect all three betas to have a positive price of risk. We find that the negative beta, which is the beta component that provides a hedge against the overall market, carries a statistically significant and economically large risk premium of 7.44% per annum. Like the total market beta, the positive beta is not statistically or economically significant. The information contained in the proposed negative and positive betas is economically and intuitively very different from the upside and downside betas in Ang et al. (2006a) and the semibetas of Bollerslev et al. (2022). The estimated price of risk associated with the negative beta is also robust to including other factors and characteristics used in the cross-sectional literature.

#17 – Momentum Effect in Anomalies/Trading Systems
#205 – Switching between Value and Momentum in Stocks
#293 – Momentum Effect in Anomalies v2
#498 – Value in Anomalies
#510 – Factor Momentum
#1021 – Timing the Factor Zoo in the US

Blitz, David: Caveats of Simple Factor Timing Strategies
https://ssrn.com/abstract=5256938
Abstract: This short article discusses the caveats of apparently simple yet highly effective factor timing strategies. Applying various such strategies to the standard U.S. Fama-French factors we find t-statistics up to 5, with combination strategies reaching even higher levels. However, some factor timing strategies have questionable after-cost profitability because of high turnover. Others suffer from substantial performance decay in recent years. Next to such general concerns we identify various strategy-specific concerns. We propose to establish minimum standards for future factor timing studies, including carefully examining explicit or implicit loadings on the phenomena described in this article, clearly establishing the distinctiveness of reported alphas, and addressing susceptibility to known limitations.

#961 – Salient Theory Predicts US Stocks in the Cross Section

Wang, Chenguang and Yao, Kai and Liu, Jinpeng: Salience, Asymmetric Effect and Stock Returns
https://ssrn.com/abstract=5234199
Abstract: Using the Method of Moments Quantile Regression (MMQR) framework, we find a strong asymmetry in the future stock return predictability of deviation salience (DS) — a measure of how a stock performs relative to its peers. Specifically, our results show negative future returns in upper quantiles and positive future returns in lower quantiles for higher DS stocks. These asymmetric effects are amplified when the limits to arbitrage are higher, such as in small, illiquid, and volatile stocks, and vary with investor sentiment and macroeconomic conditions. In addition, our findings are persistent across 47 major financial markets. Our research complements the current salience theory literature and contributes to the quantilebased asset pricing literature by showing how DS exaggerates overreactions at the tails of the return distribution.

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

Quantifying Global Real Estate Returns Over Centuries

In the realm of quantitative finance, understanding the dynamics of real estate returns over extended periods is often overlooked, which is not good, as real estate constitutes a significant portion of investors’ portfolios. The article titled Global Housing Returns, Discount Rates, and the Emergence of the Safe Asset, 1465-2024 fills the gap and provides a comprehensive historical overview of real estate yields, offering a chronological overview of real estate returns not just over a few decades but over several centuries.