Recent Academic Research
Google's weather forecasting model, Lasso regression in Korean and US markets, intra-day equity reversals, mean reversion trading the crack spread, and volume in insider trading
Welcome to this week’s report about recent academic papers on the financial markets. The newsletter has grown rapidly over the past few months, and I’m considering new features, such as implementation reports on paper strategies.
If there’s something you’d like to see, feel free to share your thoughts. And as always, filling out the weekly poll at the bottom helps me tailor content to your interests. With that said, let’s get into it.
Weather Forecasting
New, Highly Accurate Weather Forecasting Model From Google DeepMind
Paper Title: Probabilistic weather forecasting with machine learning
Authors & Date: Ilan Price, Alvaro Sanchez-Gonzalez, Ferran Alet, Tom R. Andersson, Andrew El-Kadi, Dominic Masters, Timo Ewalds, Jacklynn Stott, Shakir Mohamed, Peter Battaglia, Remi Lam, & Matthew Willson - 12/04/2024
Summary:
This paper introduces GenCast, a machine-learning-based probabilistic weather forecasting model developed by Google DeepMind. GenCast produces global 15-day ensemble forecasts at a 0.25° resolution, demonstrating superior accuracy and speed compared to the leading operational ensemble model, the European Centre for Medium-Range Weather Forecasts Ensemble Forecast System (ENS).
GenCast's improved forecasting ability is evidenced by its ability to generate sharp, realistic weather trajectories and accurately predict extreme weather events. Notably, GenCast excels in predicting extreme surface weather, including high and low temperatures, wind speeds, mean sea level pressure, tropical cyclone tracks, and regional wind power, outperforming ENS across a range of probability decision thresholds.
These findings suggest that GenCast represents a significant advancement in weather forecasting, offering potential benefits for various applications and decision-making processes.
Thoughts:
Some of you might wonder, “What does weather have to do with financial markets?” The short answer: everything. Weather impacts supply chains, energy demand, agricultural output, and even insurance claims.
An edge in forecasting extreme events - like hurricanes or typhoons - could translate into significant market advantages. For instance, accurately predicting a typhoon’s path gives traders a leg up in assessing its likely impact on regional supply chains, energy production, and specific commodities. Even smaller-scale improvements, like better rainfall or wind strength forecasts, could refine investment strategies in sectors like agriculture or renewable energy. With weather impacting countless facets of the global economy, traders with access to this model could leverage it for a significant information advantage.
FX and Equities
Lasso Regression Strategy in Korean and US Markets
Paper Title: Forecasting returns with machine learning and optimizing global portfolios: evidence from the Korean and U.S. stock markets
Authors & Date: Dohyun Chun, Jongho Kang, & Jihun Kim - 12/01/2024
Summary:
This study examines the use of machine learning to forecast returns in the Korean and U.S. stock markets. The researchers employed a variety of machine learning models, and utilized 137 financial and economic indicators as predictors. They found that Lasso and Elastic Net models consistently outperformed traditional benchmark models across all three target variables: KRW/USD exchange rate, S&P 500 returns, and KOSPI returns.
The results demonstrate that incorporating exchange rate variations into the portfolio optimization process significantly improves performance, particularly for the KOR&USFX portfolios, which invest in both Korean and U.S. market portfolios while accounting for exchange rate fluctuations.
The highest Sharpe ratios were achieved by the KOR&USFX portfolios constructed using the Lasso (Sharpe ratio = 3.45) and Elastic Net (Sharpe ratio = 3.48) models. Machine learning models outperform traditional models due to their ability to incorporate a wide variety of predictor variables that cannot be simultaneously incorporated into traditional linear models.
Thoughts:
We haven’t ventured much into foreign exchange (FX) markets here, but this paper makes a strong case for paying closer attention - especially with strategies boasting Sharpe ratios that high.
While incorporating 137 predictors into a model seems quite extreme (and makes me concerned about the potential of overfitting the data), the models perform quite well in out-of-sample testing. It’s also refreshing to see “old-school” models like LASSO and Elastic Net punching above their weight, proving you don’t always need flashy models to find alpha.
Intra-Day Reversals
“Buy the Dip”
Paper Title: End-of-Day Reversal
Authors & Date: Amar Soebhag, Guido Baltussen, & Zhi Da - 11/29/2024
Summary:
The paper explores the phenomenon of individual stock returns exhibiting a significant intraday reversal, particularly strong in the last half-hour of trading. This "end-of-day reversal" manifests as a negative relationship between a stock's returns earlier in the day and its return in the last half-hour. The reversal is primarily driven by positive price pressure on stocks that have experienced losses (negative returns) during the earlier parts of the day.
The authors find this reversal is primarily driven by stocks that experience losses during the day, and attribute it to two main factors: attention-induced retail buying and reduced short-selling activity as the trading day ends. Retail investors, drawn to large price movements, tend to purchase losing stocks, creating upward price pressure. Simultaneously, short-sellers, wary of overnight risks, decrease their positions, particularly in losing stocks, further contributing to the price reversal.
Thoughts:
We talk a lot about momentum on this newsletter, as I find the phenomenon fascinating due to the proposed behavioral explanations, like the fear of missing out (FOMO).
However, this paper’s showcased phenomenon could be exploited with more of a mean-reversion trading strategy. The “end-of-day reversal” suggests that retail investors, drawn to big intraday price drops, tend to swoop in and buy losing stocks late in the day, creating upward pressure. Add to that short-sellers unwinding their positions to avoid overnight risk, and you’ve got a recipe for a last-minute rebound.
This would be a great strategy to test in the future - let me know if you all would like to see it by filling out the poll at the bottom of the post.
Oil and Refined Products
Mean-Reversion on the Crack Spread
Paper Title: Mean Reversion Trading on the Naphtha Crack
Authors & Date: Briac Turquet, Pierre Bajgrowicz, O. Scaillet - 11/27/2024
Summary:
This paper examines short-term inefficiencies in the naphtha crack spread price, which is the price difference between naphtha - a key feedstock for petrochemical production - and crude oil, its primary input. The authors model the dynamics of standardized naphtha crack spread changes and find that mean-reversion effects are greater than transaction costs, with the majority of the mean-reversion occurring within the first day.
Through back-testing several mean-reversion trading strategies, the authors identify 7 strategies that achieve outperformance, with an average win rate of 60%. They attribute the existence of these positive returns to differences in liquidity, execution speed, and the types of participants active in the naphtha and Brent markets.
Thoughts:
Don’t be alarmed if you have never heard of the naphtha crack spread - it’s simply the price difference between the input product (Brent crude oil) and the refined product (naphtha). If you are curious, Investopedia has a helpful article explaining the crack spread in detail.
The mean-reversion abnormality likely stems from structural differences between the markets and their participants. Brent futures are traded electronically, attracting algorithmic and high-frequency traders, whereas naphtha swaps are predominantly traded through brokers by specialized market participants. Therefore, naphtha prices are slower to react to information, and allow the edge to exist.
Insider Trading
Volume is a Limiting Factor
Paper Title: Insider filings as trading signals — Does it pay to be fast?
Authors & Date: Eike Oenschläger & Steffen Möllenhoff - 11/30/2024
Summary:
This paper investigates whether swiftly reacting to insider trading announcements, specifically SEC Form 4 filings, can result in profitable trading strategies. While a simple long-stock, short-market strategy based on these filings initially shows positive abnormal percentage returns for shorter holding periods, these returns become negligible when considering realistic trading volumes.
This discrepancy arises because higher percentage returns tend to be linked with less liquid stocks, ultimately leading to lower USD returns. Even when focusing on high-volume stocks, the USD returns remain statistically insignificant, and the strategy is not scalable.
The authors conclude that a trading strategy solely based on insider trade reports is neither practical nor scalable, even before accounting for transaction costs.
Thoughts:
This paper challenges the idea that insider trading announcements are an easy source of profit. About a month ago, I highlighted research suggesting opportunities in following insider trades, but this study takes a more holistic view.
The issue? Volume. Insider trades on less liquid stocks may show higher percentage returns, but the lack of liquidity makes it difficult to convert those returns into meaningful gains. Even for high-volume stocks, the returns are not statistically significant, possibly due to market saturation from too many trades based on the same signals.
While this paper doesn’t provide a clear edge for traders (sorry!), it’s a useful reminder to always account for liquidity, transaction costs, and slippage when evaluating trading strategies.
Feedback
Thank you for reading this week’s edition of Recent Academic Research. Remember to fill out the poll to let me know which paper was your favorite and like the post if you enjoyed it. Feel free to follow up with any questions, comments, or ideas for the future!
Also, make sure to follow me on X / Twitter and BlueSky!
Disclaimer
The content provided in this newsletter, "Alpha in Academia," is for informational and educational purposes only. It should not be construed as financial advice, investment recommendations, or an offer or solicitation to buy or sell any securities or financial instruments. Past performance is not indicative of future results. The financial markets involve risks, and readers should conduct their own research and consult with qualified financial advisors before making any investment decisions.
The interpretations, opinions, and analyses presented herein are those of the author and do not necessarily reflect the views of the original researchers, their institutions, or the full implications of the cited academic papers. While every effort is made to accurately represent the research discussed, readers should be aware that the summaries and interpretations may not capture the full scope or nuances of the original studies. The information contained in this newsletter is believed to be accurate and reliable at the time of publication, but accuracy and completeness cannot be guaranteed. The author and publisher accept no liability for any loss or damage resulting from reliance on the information provided.
This newsletter may contain links to external websites or resources. The author is not responsible for the content, accuracy, or reliability of these external sources.
By subscribing to or reading this newsletter, you acknowledge that you have read and understood this disclaimer and agree to hold the author and publisher harmless from any liability that may arise from your use of the information contained herein.