Paper: arXiv 2412.11192
Authors: Igor L. R. Azevedo, Toyotaro Suzumura
Abstract
Stock market forecasting has been a topic of extensive research, aiming to provide investors with optimal stock recommendations for higher returns. In recent years, this field has gained even more attention due to the widespread adoption of deep learning models. While these models have achieved impressive accuracy in predicting stock behavior, tailoring them to specific scenarios has become increasingly important. Election Day represents one such critical scenario, characterized by intensified market volatility, as the winning candidate’s policies significantly impact various economic sectors and companies. To address this challenge, we propose the Election Day Stock Market Forecasting (EDSMF) Model. Our approach leverages the contextual capabilities of large language models alongside specialized agents designed to analyze the political and economic consequences of elections. By building on a state-of-the-art architecture, we demonstrate that EDSMF improves the predictive performance of the S&P 500 during this uniquely volatile day.
Complexity vs Empirical Score
- Math Complexity: 3.5/10
- Empirical Rigor: 7.5/10
- Quadrant: Street Traders — empirical and applied, lower math complexity
Why this score: The paper introduces advanced architectural extensions (LLM agents, political signal integration) but relies on existing frameworks with moderate mathematical depth, while demonstrating strong empirical rigor through a specific high-frequency backtest on recent election data with detailed feature engineering and ablation studies.
Research Flowchart
flowchart TD
A["Research Goal:<br>Predict S&P 500 on Election Day"] --> B["Data Sources"]
B --> C["Data Inputs:<br>Historical Price + Election News + Volatility Indices"]
C --> D["Methodology:<br>EDSMF Model Architecture"]
subgraph D ["Computational Process"]
D1["Large Language Models<br>for Contextual Analysis"]
D2["Specialized Agents<br>for Political/Economic Impact"]
end
D --> E["Model Training & Optimization"]
E --> F["Key Outcomes:<br>Improved Predictive Accuracy"]
F --> G["Impact:<br>Enhanced Volatility Forecasting"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes the Election Day Stock Market Forecasting (EDSMF) model, which integrates political signals via a large language model into a stock forecasting architecture.
Source sentence
we propose the Election Day Stock Market Forecasting (EDSMF) Model.
Assumes 4
- The model predicts the closing price of the next minute (or next day if market closed) to compute a 1‑data‑point return ratio. data
Source sentence
The task involves predicting the closing price of the next data point, which corresponds to the next minute during market hours or the next day if the market has closed.
- High‑frequency trading at a 1‑minute frequency is feasible for all S&P 500 stocks during the election period. execution
Source sentence
Unlike StockMixer, which operates at a daily frequency, our model targets high-frequency trading, focusing on all S&P 500 stocks during the period from 30/10/2024 to 06/11/2024 (U.S. election results). With a 1-minute dataset frequency, our model enables a more detailed analysis of stock movements.
- Candidate impact variables (candidate_impact_1, candidate_impact_2) accurately quantify market perception of each candidate’s policies. model
Source sentence
The candidate impact variables, candidate_impact_1 and candidate_impact_2, quantify the expected market impact of each candidate based on their proposed economic policies and campaign momentum.
- Ensemble weights (e.g., w1=0.2, w2=0.8) reflect the relative influence of the two candidates on the market during the test period. model
Source sentence
The ensemble weights reflect the importance of each candidate in influencing the market during the test period. For instance, in the provided code, the weights are set as: w1 = 0.2, w2 = 0.8.
Data
S&P 500 stocks · equities · daily
- 1‑minute S&P 500 dataset (election period) market-prices unknown
Source sentence
our model targets high-frequency trading, focusing on all S&P 500 stocks during the period from 30/10/2024 to 06/11/2024 (U.S. election results). With a 1-minute dataset frequency, our model enables a more detailed analysis of stock movements.
Costs
Not modeled
Finds 3
Sharpe 1.82
- EDSMF improves predictive accuracy and risk‑adjusted returns compared with the baseline StockMixer framework. predictive-accuracy
Source sentence
The results show that the EDSMF model improves predictive accuracy and risk-adjusted returns.
- Including political signals enables the model to outperform StockMixer on Information Coefficient (IC), Rank Information Coefficient (RIC), Precision@10, and Sharpe Ratio (SR). performance
Source sentence
The inclusion of political signals enabled the model to outperform the baseline Stock-Mixer framework in key metrics, such as Information Coefficient (IC), Rank Information Coefficient (RIC), Precision@10, and Sharpe Ratio (SR).
- Ablation studies indicate that a balanced candidate‑weight configuration (e.g., 40% Candidate‑1, 60% Candidate‑2) yields the highest Sharpe Ratio. performance
Source sentence
The configuration ‘40-Candidate-1, 60-Candidate-2‘ demonstrated the highest Sharpe Ratio, highlighting its superior risk-adjusted returns.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 24 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (22)
- StockMixer: A Simple Yet Strong MLP-Based Architecture for Stock Price Forecasting (2024) — link
- Efficient Integration of Multi-Order Dynamics and Internal Dynamics in Stock Movement Prediction (2022) — link
- HIST: A Graph-based Framework for Stock Trend Forecasting via Mining Concept-Oriented Shared Information (2021) — link
- Stock Selection via Spatiotemporal Hypergraph Attention Network: A Learning to Rank Approach (2021) — link
- Modeling the Stock Relation with Graph Network for Overnight Stock Movement Prediction (2020) — link
- HATS: A Hierarchical Graph Attention Network for Stock Movement Prediction (2019) — link
- Temporal Relational Ranking for Stock Prediction (2018) — link
- Deep learning with long short-term memory networks for financial market predictions (2017) — link
and 14 more
- Capital Asset Pricing Model (2017) — link
- CatBoost: unbiased boosting with categorical features (2017)
- XGBoost: A Scalable Tree Boosting System (2016) — link
- Why the Monte Carlo method is so important today (2014) — link
- K-nearest neighbor (2009) — link
- Stock Market Volatility around National Elections (2008) — link
- Support Vector Machines (2008) — link
- Presidential Elections and the Stock Market: Comparing Markov-Switching and Fractionally Integrated GARCH Models of Volatility (2004) — link
- Long Short-Term Memory (1997) — link
- A DISTANCE MEASURE FOR CLASSIFYING ARIMA MODELS (1990) — link
- Generalized autoregressive conditional heteroskedasticity (1986) — link
- Stock Market Returns and the Presidential Election Cycle: Implications for Market Efficiency (1980) — link
- Portfolio Selection (1971) — link
- From Votes to Volatility Predicting
0 of 22 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.