Long-short Trading Strategy Research Articles

Firm connection and equity return predictability – Graph-based machine learning methods

We develop a unified measure of firm linkage from popular linkage indicators in the literature via graph-based machine learning methods and investigate its asset pricing implication. Using all A-shares listed in the Chinese stock market from 2003 to 2022, we reveal a widespread momentum spillover in the cross section of stock returns based on our linkage measure. In particular, a long-short trading strategy for portfolios sorted by our measure generates significant risk-adjusted returns of 0.83% on a monthly basis. We show that our measure contains incremental information on firm fundamental connections relative to well-documented alternative measures, and its predictive ability can be rationalized by the investor inattention hypothesis. Our study contributes to the literature which explores economic linkages that generate lead-lag predictability and sheds new light on the interconnected nature of companies in an economy via advanced machine learning techniques.

VIX constant maturity futures trading strategy: A walk-forward machine learning study.

This study employs seven advanced machine learning approaches to conduct numerical predictions of the next-day returns of VIX constant-maturity futures (VIX CMFs) using the term structure information derived from VIX CMFs. Based on precise numerical predictions, this study proposes a new Constrained-Mean-Variance Portfolio Optimization (C-MVO) trading strategy and tests it against a benchmark long-short trading strategy to evaluate the profitability of the machine learning numerical predictions. This study applies three unique feature sets, each incrementally incorporating the VIX CMFs' term structure features, to individually examine the predictive ability of the seven machine learning models and their backtesting performance. Over a comprehensive 11-year period, the experiment adheres to a strict walk-forward expanding-window methodology for both training and backtesting. The predictive and backtesting results show that four of the seven machine learning models attain a prediction information ratio greater than 0.02, with an average prediction information ratio of 0.037. This result suggests that the VIX CMFs term structure features have predictive power for the next-day returns of VIX CMFs. Moreover, the average C-MVO information ratio is 0.623, and the long-short strategy information ratio is 0.404. This increase in the information ratio under the C-MVO strategy validates the effectiveness of the machine learning models and the C-MVO strategy.

Can convertible bond trading predict stock returns? Evidence from China

This paper studies the information content of convertible bond trading in Chinese market. We construct a measure of the convertible bond order imbalance. The analysis shows that the convertible bond order imbalance positively and significantly predicts future stock returns. The results hold in both daily and 30-min frequencies. Our findings are robust by considering the stock price sensitivity of convertible bond. The predictability is of economic significance that the long-short trading strategy based on the convertible bond order imbalance generates an annual return of 17.54%. Moreover, the predictive relations varies with moneyness, conversion premium, limits to arbitrage, level of information asymmetry and maturity. Since stocks are subject to “T + 1” trading rule while convertible bonds are not in Chinese market, convertible bond market provides an alternative venue to gain stock exposures with less trading restrictions. We find that the stock order imbalance exhibits a weaker predictability on stock returns compared to the convertible bond order flows. Our findings suggest that convertible bond market plays an important role in discovering private information of stock prices.

The countercyclical long-term operating accrual-based trading strategy in the Stoxx Europe 600 index: The importance of asset and liability components.

This work uses long-term operating accruals, rather than current, as an accounting measure to identify major anomalies. Past and abundant accounting and financial literature associates anomalies with problems of reliability and assigns lower reliability to long-term operating accruals than to current accruals. We investigate the relation between scaled operating accruals and size-adjusted abnormal returns for nonfinancial firms listed in the Stoxx Europe 600 index for the period 2000–2021. We find consistent evidence of (1) a higher long-term operating accrual anomaly than working capital accrual, especially, when asset and liability components are separated (2) long-short trading strategies aimed at taking advantage of the anomaly that achieves significant annual returns between 2% and 6% and (3) this trading strategy strongly reduces the risk of stock portfolios during an economic crisis due to its countercyclical nature. These findings have important implications not only for academics, but also for asset managers who want to protect the return of their stock portfolios from high market volatility.

Cryptocurrency News Website with Prediction

We analyze the predictability of the bitcoin market across prediction horizons ranging from 1 to 60 min. In doing so, we test various machine learning models and find that, while all models outperform a random classifier, recurrent neural networks and gradient boosting classifiers are especially well-suited for the examined prediction tasks. We use a comprehensive feature set, including technical, blockchain-based, sentiment-/interest-based, and asset-based features. Our results show that technical features remain most relevant for most methods, followed by selected blockchain- based and sentiment-/interest-based features. Additionally, we find that predictability increases for longer prediction horizons. Although a quantile-based long-short trading strategy generates monthly returns of up to 39% before transaction costs, it leads to negative returns after taking transaction costs into account due to the particularly short holding periods. Cryptocurrencies, which the Bitcoin is the most remarkable one, have allured substantial awareness up to now, and they have encountered enormous instability in their price. While some studies utilize conventional statistical and econometric ways to uncover the driving variables of Bitcoin's prices, experimentation on the advancement of predicting models to be used as decision support tools in investment techniques is.

Uncovering feature interdependencies in high-noise environments with stepwise lookahead decision forests

Conventionally, random forests are built from “greedy” decision trees which each consider only one split at a time during their construction. The sub-optimality of greedy implementation has been well-known, yet mainstream adoption of more sophisticated tree building algorithms has been lacking. We examine under what circumstances an implementation of less greedy decision trees actually yields outperformance. To this end, a “stepwise lookahead” variation of the random forest algorithm is presented for its ability to better uncover binary feature interdependencies. In contrast to the greedy approach, the decision trees included in this random forest algorithm, each simultaneously consider three split nodes in tiers of depth two. It is demonstrated on synthetic data and financial price time series that the lookahead version significantly outperforms the greedy one when (a) certain non-linear relationships between feature-pairs are present and (b) if the signal-to-noise ratio is particularly low. A long-short trading strategy for copper futures is then backtested by training both greedy and stepwise lookahead random forests to predict the signs of daily price returns. The resulting superior performance of the lookahead algorithm is at least partially explained by the presence of “XOR-like” relationships between long-term and short-term technical indicators. More generally, across all examined datasets, when no such relationships between features are present, performance across random forests is similar. Given its enhanced ability to understand the feature-interdependencies present in complex systems, this lookahead variation is a useful extension to the toolkit of data scientists, in particular for financial machine learning, where conditions (a) and (b) are typically met.

Stock price manipulation, short-sale constraints, and breadth-return relationship

We examine the breadth-return relationship and the effect of stock price manipulation on it. Using data from the Chinese stock market, we first empirically find a significantly negative breadth-return relationship. Moreover, we conduct a long-short trading strategy based on breadth change, and show that the magnitude of long-short hedged portfolio based on easily-manipulated stocks is at least 46% more than that based on difficultly-manipulated stocks, suggesting that the negative breadth-return relationship is more pronounced for stocks that are easier to be manipulated. Different from the popularity view of Choi et al. (2013), our results support the “manipulation view” that manipulation behaviors of large investors drive the negative breadth-return relationship. Furthermore, the negative breadth-return relationship is alleviated after relaxing short-sale constraints, indicating that the short-sale mechanism can help reduce stock price manipulation. This paper sheds new light on the role of ownership breadth.

Short-term bitcoin market prediction via machine learning

We analyze the predictability of the bitcoin market across prediction horizons ranging from 1 to 60 min. In doing so, we test various machine learning models and find that, while all models outperform a random classifier, recurrent neural networks and gradient boosting classifiers are especially well-suited for the examined prediction tasks. We use a comprehensive feature set, including technical, blockchain-based, sentiment-/interest-based, and asset-based features. Our results show that technical features remain most relevant for most methods, followed by selected blockchain-based and sentiment-/interest-based features. Additionally, we find that predictability increases for longer prediction horizons. Although a quantile-based long-short trading strategy generates monthly returns of up to 39% before transaction costs, it leads to negative returns after taking transaction costs into account due to the particularly short holding periods.

Common dynamic factors for cryptocurrencies and multiple pair-trading statistical arbitrages

In this paper, we apply dynamic factor analysis to model the joint behaviour of Bitcoin, Ethereum, Litecoin and Monero, as a representative basket of the cryptocurrencies asset class. The empirical results suggest that the basket price is suitably described by a model with two dynamic factors. More precisely, we detect one integrated and one stationary factor until the end of August 2019 and two integrated factors afterwards. Based on this evidence, we define a multiple long-short trading strategy which proves profitable when the second factor is stationary.

Deep Reinforcement Learning on a Multi-Asset Environment for Trading

Financial trading has been widely analyzed for decades with market participants and academics always looking for advanced methods to improve trading performance. Deep reinforcement learning (DRL), a recently reinvigorated method with significant success in multiple domains, still has to show its benefit in the financial markets. We use a deep Q-network (DQN) to design long-short trading strategies for futures contracts. The state space consists of volatility-normalized daily returns, with buying or selling being the reinforcement learning action and the total reward defined as the cumulative profits from our actions. Our trading strategy is trained and tested both on real and simulated price series and we compare the results with an index benchmark. We analyze how training based on a combination of artificial data and actual price series can be successfully deployed in real markets. The trained reinforcement learning agent is applied to trading the E-mini S&P 500 continuous futures contract. Our results in this study are preliminary and need further improvement.

Climate Change Exposure and Stock Return Predictability

This paper finds evidence that stock returns vary with the physical climate change exposure of firms in a predictable manner. We construct measures of exposures to physical climate changes at the firm level, and find that firms with high climate change exposures experience lower future profitability. We show that due to limited attention of investors, stock prices do not promptly incorporate such climate change information, and firms with high climate change exposures are subject to subsequent lower returns. A long-short trading strategy based on this effect produces significant alphas of around 0.6% per month. Moreover, we document that our return predictability strengthens under severe physical climate conditions, and it is not the result of risks associated with carbon emissions.

Cross-Predictability of Earnings and Stock Returns: A Latent Factor Approach

I propose a new method, three-pass regression filter developed by Kelly and Pruitt (2015), to predict non-announcing firms’ earnings news using the cross-section of all available early announcers’ earnings news, the number of which can be as large as thousands. The method assumes a set of common latent factors driving the earnings news of non-announcing firms and early announcers and thus efficiently reduces the dimension of announced earnings. Empirical tests show that the extracted measure strongly predicts the earnings surprise and the earnings announcement return with both statistical and economic significance. A long-short trading strategy based on the extracted information realizes a 10% value-weighted alpha annually, indicating a delayed reaction of investors. The return predictability is stronger for small firms or firms with less investor attention. Controlling a series of documented information channels has little impact on the predictive power of this extracted measure.

The V-Shaped Disposition Effect in the Stock Exchange of Thailand

The objective of this study is to investigate how investors in the Stock Exchange of Thailand practically trade in response to a magnitude of profits and losses, given a discussion of the widely well-known behavioral explanation, so called as the disposition effect. We provide empirical evidence of an existence of the V-shaped disposition effect, which has been recently found in several advanced equity markets. By adopting the methodology suggested by An’s (2016) and Fama and Macbeth (1973), we document that stock return patterns in relation to aggregate unrealized gains and losses of investors are consistent with the V-shaped selling schedule, given an increase in unrealized gains and losses over the period of January 1996 to December 2015. The effect of unrealized gains is stronger than that of unrealized losses and this asymmetry underlies the existence of the V-shaped disposition effect in the Thai equity market. Interestingly, the effect of the V-shaped selling schedule is strongest over the short-term holding time horizon. Last but not the least, stocks for which investors have large unrealized gains and losses outperform in the following month and the long-short trading strategy, based on this premise, generates the average 1.7% monthly (equivalent to 20.0% per year) abnormal return.

Investor Base Size and Stock Return Anomalies

We find that several well-documented stock return anomalies arise and persist only among stocks with smaller (institutional) investor bases, which are presumably stocks that are neglected by investors. These results are driven by the short side of the long-short trading strategies, they appear even after controlling for several stock characteristics (e.g., market capitalization and institutional ownership) and potential risk factors, and they are considerably more pronounced during periods with more information and/or less technology. Overall, these findings suggest that the incomplete dissemination of (negative) information across investors helps in explaining the occurrence and the persistence of cross-sectional stock return anomalies.

The Information Content of Target Price Forecasts: Evidence from Mergers and Acquisitions

This paper investigates the informativeness and value relevance of analyst target prices in the context of mergers and acquisitions (M&A). Our results indicate that firms with high 12-month ahead target prices relative to current stock prices are more likely to become a takeover target and offer premium and acquirers’ announcement returns are positively associated with firms’ target price premium. We also show that a long-short trading strategy formed on target prices and firms’ takeover likelihood generates economically significant returns. Our results are robust to a battery of additional analysis, and the informativeness of target prices is not subsumed by other analyst-forecast outputs such as earnings forecasts and recommendations. Overall, our findings suggest that analysts convey valuable information through target price forecasts, which are useful for participants in the corporate takeover markets.

Show me the money: The monetary policy risk premium

We create a parsimonious monetary policy exposure (MPE) index based on observable firm characteristics that previous studies link to how stocks react to monetary policy. Our index successfully captures stocks’ responses to both conventional and unconventional monetary policy. Stocks whose prices react more positively to expansionary monetary policy (high-MPE stocks) earn lower average returns. This result is consistent with the notion that high-MPE stocks provide a hedge against bad economic shocks, to which the Federal Reserve responds with expansionary monetary policy. A long-short trading strategy designed to exploit this effect achieves an annualized Sharpe Ratio of 0.77.

EMD2FNN: A strategy combining empirical mode decomposition and factorization machine based neural network for stock market trend prediction

Stock market forecasting is a vital component of financial systems. However, the stock prices are highly noisy and non-stationary due to the fact that stock markets are affected by a variety of factors. Predicting stock market trend is usually subject to big challenges. The goal of this paper is to introduce a new hybrid, end-to-end approach containing two stages, the Empirical Mode Decomposition and Factorization Machine based Neural Network (EMD2FNN), to predict the stock market trend. To illustrate the method, we apply EMD2FNN to predict the daily closing prices from the Shanghai Stock Exchange Composite (SSEC) index, the National Association of Securities Dealers Automated Quotations (NASDAQ) index and the Standard & Poor’s 500 Composite Stock Price Index (S&P 500), which respectively exhibit oscillatory, upward and downward patterns. The results are compared with predictions obtained by other methods, including the neural network (NN) model, the factorization machine based neural network (FNN) model, the empirical mode decomposition based neural network (EMD2NN) model and the wavelet de-noising-based back propagation (WDBP) neural network model. Under the same conditions, the experiments indicate that the proposed methods perform better than the other ones according to the metrics of Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE). Furthermore, we compute the profitability with a simple long-short trading strategy to examine the trading performance of our models in the metrics of Average Annual Return (AAR), Maximum Drawdown (MD), Sharpe Ratio (SR) and AAR/MD. The performances in two different scenarios, when taking or not taking the transaction cost into consideration, are found economically significant.

A Generalized Framework for Simultaneous Long-Short Feedback Trading

We present a generalization of the simultaneous long-short (SLS) trading strategy described in recent control literature wherein we allow for different parameters across the short and long sides of the controller; we refer to this new strategy as generalized SLS (GSLS). Furthermore, we investigate the conditions under which positive gain can be assured within the GSLS setup for both deterministic stock price evolution and geometric Brownian motion. In contrast to existing literature in this area (which places less emphasis on the practical application of SLS strategies), we suggest optimization procedures for selecting the control parameters based on historical data, and we extensively test these procedures across a large number of real stock price trajectories (495 in total). We find that the implementation of such optimization procedures greatly improves the performance compared with fixing control parameters, and, indeed, the GSLS strategy outperforms the simpler SLS strategy in general.

Trading on ETF mispricings

PurposePrior literature has shown deviations between ETF prices and their net-asset-value (NAV) to exist. Fulkerson and Jordan (2013, p. 31) question “if there exists a true tradeable strategy” to exploit these inefficiencies. The purpose of this paper is to implement a profitable daily long-short trading strategy based on price/NAV information and explicitly accounting for trading costs.Design/methodology/approachFor a sample of European sector ETFs, the authors analyze gross and net returns of a long-short trading strategy in the capital asset pricing model and Fama-French three-factor model.FindingsThe authors document positive gross excess return for the long-short trading strategy in all sample periods, but net excess returns to be positive only between 2008 and 2010.Research limitations/implicationsThe results document a profitable long-short trading strategy exploiting deviations between ETF prices and NAV and highlight the impact of trading costs in ETF markets. Due to the limited availability of historic trading cost data, the research uses a comparatively small sample size.Practical implicationsThe net profitability of long-short trading in ETFs is only found in times of high uncertainty in the stock market.Originality/valueThe inclusion of trading costs enables a detailed comparison between gross and net returns in ETF trading, addressing potential limits to arbitrage.

Investor recognition and stock returns: evidence from China

PurposeMerton’s model of capital market equilibrium under incomplete information predicts that contemporaneous stock returns are positively related to investor recognition and that future stock returns are negatively related to investor recognition. The purpose of this paper is to empirically investigate whether Merton’s theory holds true for the Chinese stock market.Design/methodology/approachThis paper proposes the degree of shareholder base growth (SBG) as a proxy for investor recognition and examines the relationship between investor recognition and stock returns through a univariate analysis and Fama-Macbeth cross-sectional regressions based on A-Share listed firms.FindingsThe results show that investor recognition is nonlinearly and positively related to contemporaneous stock returns and is negatively related to future stock returns in contrast to the conclusions of Merton’s theory. A long-short trading strategy that involves buying stocks with the lowest SBG rate and that sells stocks with the highest SBG rate will earn an average monthly return of 3.615 percent.Research limitations/implicationsThough Merton’s theory is not fully reflected in the Chinese stock market, investor recognition is considered an important risk factor in the Chinese stock market.Originality/valueNo works have yet investigated the validity of Merton’s “investor cognition hypothesis” in relation to the Chinese stock market. This paper strives to fill this gap.