Quantpedia Premium Update – 1st February 2022

New strategies:

#713 –Synthetic Lending Rates Predict Subsequent Market Return

Period of rebalancing: Daily
Markets traded: equities
Instruments used for trading: ETFs
Complexity: Moderately complex strategy
Backtest period: 2016-2021
Indicative performance: 15.47%
Estimated volatility: 19.52%

Source paper:

Padyšák, Matúš: Synthetic lending rates predict subsequent market return
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3976307
Abstract:
The paper studies the relationship between synthetic lending rates derived from the options market and subsequent market performance. The research examines Cboe Hanweck Borrow Intensity Indicators, defined as the risk-free rate minus the lending fee. According to the results, the rise (fall) in aggregate borrow intensity computed as the average borrow intensity for more than 4000 assets predicts positive (negative) next day’s market return proxied by the SPY ETF. The effect is statistically and economically significant but is quickly reversed over the next two days. Additionally, the effect is much more substantial during crisis periods in the sample: the crash of December 2018 and the beginning of the coronavirus pandemic (February – April 2020). The two crises are the main reasons the market timing strategiesoutperform the SPY benchmark during the sample period. Overall, the results show crucial implications of changes in borrow intensities and lending fees during crises.

#714 –Illiquidity Factor in China

Period of rebalancing: Monthly
Markets traded: equities
Instruments used for trading: stocks
Complexity: Simple strategy
Backtest period: 1997-2019
Indicative performance: 115.32%
Estimated volatility: 33.93%

Source paper:

Jun Liua, Kai Wua, Fuwei Jianga: How Is Illiquidity Priced in the Chinese Stock Market?
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3787113
Abstract:
Using 3,589 stocks in China, we find that the liquidity premium ranges from 3.3% to 8.0% per month, and it remains to persist after controlling other well-known firm characteristics. The liquidity-based strategy explains China’s cross-section and time-series expected returns regardless of which measure is employed. It suggests that investors in China are faced with high transaction costs and systematic risk. Our multivariate decomposition approach highlights that characteristic momentum and idiosyncratic volatility account for more than 60% of the illiquidity premium.

#715 –Investment Effect in China

Period of rebalancing: 6 months
Markets traded: equities
Instruments used for trading: stocks
Complexity: Moderately complex strategy
Backtest period: 1998-2019
Indicative performance: 10.95%
Estimated volatility: 9.38%

Source paper:

Huixuan Li, Jing Che: Does Higher Investments Necessarily Reduce Stock Returns?
https://ssrn.com/abstract=3954124
Abstract:
Previous studies find corporate investments negatively predict firm performance and stock returns. Using data from the Chinese A-share stock market, we find firms that substantially increase their investments have higher, rather than lower, subsequent stock returns. This effect persists after controlling for important price factors such as size, value, momentum, and turnover. Further analysis shows that this effect is more pronounced among large firms, low BM firms, profitable firms and high-growth industries. We offer a unified explanation for the contradictory findings: The investment-stock return relation depends on whether return to capital is decreasing or increasing. We find evidence of increasing returns to capital in last two decades in China, which implies more investment leads to higher rather than lower profits and consequently higher stock returns.

#716 –Accruals Seasonality

Period of rebalancing: Monthly
Markets traded: equities
Instruments used for trading: stocks
Complexity: Moderately complex strategy
Backtest period: 2001-2019
Indicative performance: 7.28%
Estimated volatility: 8.26%

Source paper:

Siu Kai Choy, Gerald J. Lobo, Yongxian Tan: The Rise of Accruals Seasonality Spread
https://ssrn.com/abstract=3983051
Abstract:
We document that historical patterns of accruals seasonality predict future stock returns. Firms with historically larger accruals in a given quarter of the year earn lower stock returns when those accruals are expected to be announced. The accruals seasonality spread is significant only in the post-2001 period when there are increased accruals-based arbitrage activities. Further analysis indicates that the accruals seasonality anomaly is not explained by risk-based explanations, but results from unsophisticated arbitrage of the accruals anomaly.

New research papers related to existing strategies:

#253 – Cross-Sectional Momentum in Futures
#297 – Combining Time-Series and Cross-Sectional Momentum
#342 – Global Cross-Asset Time Series Momentum in Bond and Equity Markets

Pitkäjärvi, Aleksi: Decomposing Cross-Asset Time Series Momentum
https://ssrn.com/abstract=3714611
Abstract:
I decompose the expected return difference between cross-asset time series momentum and time series momentum into market timing and risk premium components, and show that market timing accounts for 71–79% of the difference. I thus show that two recent critiques of time series momentum do not apply to cross-asset time series momentum. Instead, the outperformance of cross-asset time series momentum is driven specifically by the strategy’s ability to exploit cross-asset time series predictability in global bond and equity markets.

#685 – Boosted Trees and Cryptocurrency Return Prediction

Jakubik, Johannes and Nazemi, Abdolreza and Geyer-Schulz, Andreas and Fabozzi, Frank J.: Incorporating Financial News for Forecasting Bitcoin Prices Based on Long Short-Term Memory Networks
https://ssrn.com/abstract=3733398
Abstract:
In this paper we investigate how a deep learning machine learning model can be applied to improve Bitcoin price forecasting and trading by incorporating unstructured information from financial news. The two-stage model we propose outperforms other machine learning models significantly. In the first stage, we leverage long short-term memory (LSTM) networks to extract structured information from financial news. In the second stage, we apply a second LSTM network with structured input from financial news to the prediction of Bitcoin prices. In addition to the superior performance relative to other machine learning models, we find that the out-of-time rate of return attained with the proposed deep learning model is substantially higher than for a buy-and-hold strategy. Our study highlights how combining deep learning and financial news offers investors and traders support for the monetization of unstructured data in finance.

#529 – Identifying Smart Money with Options

Wang, Li and Ni, Sophie Xiaoyan and Stouraitis, Aristotelis: Index Options Trading and Sentiment
https://ssrn.com/abstract=3981994
Abstract:
We find public directional order imbalance of in-the-money options on S&P500 (“DOI”) negatively predicts market returns over a monthly horizon, up until three months, after controlling for established sentiment measures and market return predictors. Stocks in the High-tech industry are the most sensitive to DOI, consistent with findings in sentiment literature. We also find that the predictability of the DOI is stronger in the recession period, supporting limits to arbitrage hypothesis. Evidence has shown that DOI predicts market return through both cash flow and discount rate channels. These results provide a new piece of evidence on investors’ irrational trading motivation for index options.

#670 – Machine Learning Pairs Trading Strategy

Sang-Ho Kim, Deog-Yeong Park and Ki-Hoon Lee: Hybrid Deep Reinforcement Learning for Pairs Trading
https://www.mdpi.com/2076-3417/12/3/944/pdf
Abstract:
Pairs trading is an investment strategy that exploits the short-term price difference (spread) between two co-moving stocks. Recently, pairs trading methods based on deep reinforcement learning have yielded promising results. These methods can be classified into two approaches: (1) indirectly determining trading actions based on trading and stop-loss boundaries and (2) directly determining trading actions based on the spread. In the former approach, the trading boundary is completely dependent on the stop-loss boundary, which is certainly not optimal. In the latter approach, there is a risk of significant loss because of the absence of a stop-loss boundary. To overcome the disadvantages of the two approaches, we propose a hybrid deep reinforcement learning method for pairs trading called HDRL-Trader, which employs two independent reinforcement learning networks; one for determining trading actions and the other for determining stop-loss boundaries. Furthermore, HDRL-Trader incorporates novel techniques, such as dimensionality reduction, clustering, regression, behavior cloning, prioritized experience replay, and dynamic delay, into its architecture. The performance of HDRL-Trader is compared with the state-of-the-art reinforcement learning methods for pairs trading (P-DDQN, PTDQN, and P-Trader). The experimental results for twenty stock pairs in the Standard & Poor’s 500 index show that HDRL-Trader achieves an average return rate of 82.4%, which is 25.7%P higher than that of the second-best method, and yields significantly positive return rates for all stock pairs.

#441 – Profitability Factor Combined with Value Factor

Black, Stanley and Dai, Wei: Assessing the Relative Magnitude of Premiums
https://ssrn.com/abstract=3981766
Abstract:
We examine the magnitude of the expected size, value, and profitability premiums relative to each other and across regions using monthly premiums from 1927 to 2020. We do not find reliable differences in expected premiums, individually or jointly, across the US, developed ex US, and emerging markets. We also do not find reliable differences across premiums within a region. As a result, we believe that investors should be cautious about favoring one premium over another or one region over another based on the magnitude of the expected premiums.

#536 – Machine Learning Stock Picking

Binz, Oliver and Schipper, Katherine and Standridge, Kevin: What Can Analysts Learn from Artificial Intelligence about Fundamental Analysis?
https://ssrn.com/abstract=3745078
Abstract:
We use machine learning to estimate Nissim and Penman’s (2001) (NP) structural framework that decomposes profitability into four levels of increasingly disaggregated profitability drivers. Our analysis has two distinct features: first, we apply machine learning to accommodate the non-linearities that precluded NP from estimating their framework and second, we analyze the financial statement design choices in NP to provide insights for the teaching and practice of fundamental analysis. We find that out-of-sample profitability forecasts obtained by applying machine learning to NP’s framework are more accurate than those from benchmark models, and that investing strategies based on intrinsic values generated from our profitability forecasts yield risk-adjusted returns. With respect to insights for fundamental analysis, we find that focusing on operating activities, core items and five-year-horizon forecasts improves performance while using a long time series of past information impairs performance. We find mixed evidence of benefits from increasingly granular disaggregation of profitability.

#57 – Term Spread Premium

Choi, Jaehyuk and Ge, Desheng and Kang, Kyu H. and Sohn, Sungbin: Yield Spread Selection in Predicting Recession Probabilities: A Machine Learning Approach
https://ssrn.com/abstract=3723717
Abstract:
The literature on using yield curves to forecast recessions customarily uses 10-year–three-month Treasury yield spread without verification on the pair selection. This study investigates whether the predictive ability of spread can be improved by letting a machine learning algorithm identify the best maturity pair and coefficients. Our comprehensive analysis shows that, despite the likelihood gain, the machine learning approach does not significantly improve prediction, owing to the estimation error. This is robust to the forecasting horizon, control variable, sample period, and oversampling of the recession observations. Our finding supports the use of the 10-year–three-month spread.

#50 – FED Model

Asness: The FED model and expected asset returns
http://www.ebs.edu/fileadmin/redakteur/funkt.dept.finance/DGF/downloads/Paper/No-164.pdf
Abstract:
The earnings yield and long-term bond yields have been widely used to predict asset returns. In this paper, I focus on the predictive role of the stock-bond “yield gap” – the difference between the earnings yield and the 10 year Treasury bond yield – also know as the FED model, and which can be interpreted as a long term yield spread of stocks relative to bonds. Conditional on other forecasting variables, the yield gap forecasts positive excess stock returns, both at short and long forecasting horizons, although the forecasting power is greater at the near horizons. On the other hand, the yield gap forecasts negative excess returns for bonds, at both short and long horizons. A VAR variance decomposition for stock market returns, shows that shocks in the yield gap are highly positively correlated with innovations in both future discount-rate and cash flow news, confirming that the spread conveys information about future earnings and returns. An investment strategy based on the forecasting ability of the Yield gap produces higher Sharpe ratios than passive strategies in both the market index and long-term bond. In the context of an equilibrium multifactor ICAPM, the yield gap has some explanatory power over the cross section of stock returns.

#187 – CEO Interviews Effect

Flam, Rachel and Green, Jeremiah and Sharp, Nathan Y.: Do Investors Respond to CEO Facial Expressions of Anger During Television Interviews?
https://ssrn.com/abstract=3740755
Abstract:
Televised media interviews with public company CEOs occur nearly every trading day. During these interviews, investors observe visual cues in addition to hearing the verbal information managers disclose. Building on findings in the psychology and communications literature, we ask whether investors learn from CEO facial expressions. Using a sample of 959 interviews on CNBC from 2014-2018, we focus on CEO expressions of anger, an emotion generally associated with negative outcomes. We find that CEOs are more likely to show facial expressions of anger when the CEO is more expressive generally, when the journalist shows an angry facial expression, and when recent stock returns are lower. We also find that investors respond negatively to CEO facial expressions of anger and that CEO anger can nullify the benefits of a positive message from journalists.

#88 – 52-Weeks High Effect in Stocks

Hong, Jordan, Liu: Industry Information and the 52-Week High Effect
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1787378
Abstract:
We find that the 52-week high effect (George and Hwang, 2004) cannot be explained by risk factors. Instead, it is more consistent with investor under-reaction caused by anchoring bias: the presumably more sophisticated institutional investors suffer less from this bias and buy (sell) stocks close to (far from) their 52-week highs. Further, the effect is mainly driven by investor under-reaction to industry instead of firm-specific information. The extent of under-reaction is more for positive than for negative industry information. A strategy that buys stocks in industries in which stock prices are close to 52-week highs and shorts stocks in industries in which stock prices are far from 52-week highs generates a monthly return of 0.60% from 1963 to 2009, roughly 50% higher than the profit from the individual 52-week high strategy in the same period. The 52-week high strategy works best among stocks with high R-squares and high industry betas (i.e., stocks whose values are more affected by industry factors and less affected by firm-specific information). Our results hold even after controlling for both individual and industry return momentum effects.

#4 – Overnight Anomaly

Kelly, Michael A.: Overnight Returns as a Market Timing Strategy
https://ssrn.com/abstract=3692068
Abstract:
Risk-adjusted overnight returns greatly exceed risk-adjusted daytime returns. Researchers use Jensen’s alpha or the Fama-French three factor model for risk adjustment and use Fama-MacBeth regressions to test the estimated betas’ predictivity. However, owning stocks only during the day or night is a market timing strategy. Using the non-linear factors proposed by Merton (1981) and Goetzmann et al. (2000), we show that the close-to-open strategy has negative market timing ability (measured by a non-linear regressor) with a positive selectivity alpha (measured by alpha), while the open-to-close strategy has the opposite. We also find that alpha is significant in down market periods.

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

VIX-Yield Curve Cycles May Predict Recessions

Since recessions and bear markets come hand in hand for several asset classes, recession predictions have always been the foremost concern. The yield curve slope, defined as the difference between long and short-term rates, is the leading indicator backed by numerous research papers. Hansen (2021) builds on this theorem, but the author improves the recession prediction by his empirical observation that the VIX index (index of implied equity volatility or fear index) and the slope co-move in counterclockwise cycles, which align with business cycles.

Factor Performance in Bull and Bear Markets

Do common equity factors suffer during bear markets? Undoubtedly, the market factor is a rather unpleasant investment during bear markets, but what about the long-short factors? Are they able to deliver performance? The research paper by Geertsema and Lu (2021) provides several answers and interesting insights.

Introduction to Dollar-Cost Averaging Strategies

Most of you have probably heard the saying that somebody “averaged” into or out of his investment position. But what does it exactly mean, and what different dollar-cost averaging strategies exist? We plan to unveil our new “Dollar-Cost Averaging” report for Quantpedia Pro clients next week, and this article serves as a short introduction to this term.