Impact of Global Events on S&P 500 Index

🎯 Overview

Understanding how global crises shape financial markets is crucial for investors, policymakers, and financial analysts. This project analyzes 23 years of S&P 500 data (2000-2024) to uncover how major historical events impact different market sectors and to predict future market behavior using advanced statistical and machine learning techniques.

Why this matters:

• 📊 For Investors: Identify sectors that demonstrate resilience during crises and optimize portfolio diversification strategies
• 🏛️ For Policymakers: Understand differential recovery patterns across industries to design targeted economic interventions
• 🔮 For Analysts: Leverage predictive models to forecast market responses to future global disruptions
• 📚 For Researchers: Access reproducible analysis combining classical econometrics (ARIMA) with modern deep learning (LSTM)

---

🔍 Research Questions

1. Which sectors drive the S&P 500 index?

   • Using ANOVA and linear regression to identify primary market drivers

2. How do different industries respond to major historical events?

   • Statistical comparison of sector performance before/after crises

3. Can we forecast the S&P 500 index effectively?

   • Comparative evaluation of ARIMA vs. LSTM models

---

🌍 Events Analyzed

The Great Recession (2007-2009)

Financial crisis triggered by subprime mortgage collapse

• Impact: Financials, Energy, and Real Estate sectors devastated
• Insight: Banking sector took 3+ years to recover

COVID-19 Pandemic (2020-2023)

Global health crisis causing unprecedented economic disruption

• Impact: Technology sector surged; Real Estate and Utilities declined
• Insight: Digital transformation accelerated by 5-7 years

Russia-Ukraine Invasion (2022-present)

Geopolitical conflict affecting global supply chains

• Impact: Energy and Industrials gained; Technology remained resilient
• Insight: Energy sector benefited from supply disruptions and pricing power

---

📊 Key Findings

Sector Resilience Rankings

Crisis	Most Resilient	Most Vulnerable
Great Recession	Health Care, Consumer Staples	Financials, Real Estate, Energy
COVID-19	Information Technology, Energy	Real Estate, Utilities
Russia-Ukraine	Information Technology, Energy	Real Estate

Cross-Event Pattern: Real Estate consistently underperformed across all three crises, while Information Technology demonstrated exceptional resilience.

Model Performance

Our analysis reveals that LSTM neural networks outperform traditional ARIMA models in forecasting accuracy:

• ARIMA: Fast computation (~minutes), moderate accuracy
• LSTM: High accuracy, captures complex patterns, requires more computational resources (~hours)

Practical Implication: For short-term trading decisions requiring rapid updates, ARIMA suffices. For strategic portfolio management, LSTM's superior accuracy justifies the computational cost.

Statistical Validation

• ✅ All sector impacts on closing prices are highly significant (p < 2.2e-16)
• ✅ T-tests confirm statistically significant changes in sector performance post-crisis (p < 0.05)
• ✅ Information Technology and Consumer Discretionary are primary index drivers

---

🛠️ Technologies

Statistical Analysis: R (≥ 4.0) with ANOVA, t-tests, time series analysis
Machine Learning: LSTM neural networks (Keras/TensorFlow)
Data: 500+ companies, 11 GICS sectors, 6,000+ trading days
Reproducibility: R Markdown notebooks with full documentation

📖 See INSTALLATION.md for setup instructions

---

🚀 Quick Start

# Clone the repository
git clone https://github.com/DATS6101-TeamNeo/final-project.git
cd final-project

# Open analysis/Main.Rmd in RStudio and knit to generate the full report

For detailed installation and usage instructions, see INSTALLATION.md

---

📁 Project Structure

├── analysis/                     # R Markdown analysis notebooks
│   ├── Main.Rmd                  # 🎯 Complete analysis (START HERE)
│   ├── Summary.Rmd               # Executive summary
│   ├── EDA.Rmd                   # Exploratory data analysis
│   └── ARIMA_baseline.Rmd        # ARIMA forecasting model
├── reports/                      # Generated HTML reports
│   ├── Main.html
│   ├── Summary.html
│   ├── EDA.html
│   └── ARIMA_baseline.html
├── data/
│   ├── raw/                      # Historical datasets
│   │   ├── sp500_companies.csv
│   │   └── sp500_index.csv
│   └── scripts/                  # Data collection scripts
│       └── Generate_SandP500.Rmd
├── models/
│   └── checkpoints.h5            # Pre-trained LSTM model
├── figures/                      # Generated plots and images
├── docs/
│   ├── INSTALLATION.md           # Setup guide
│   └── Final-Project-Proposal.pdf
├── README.md
└── LICENSE

---

💡 Practical Applications

For Portfolio Managers

• Diversification Strategy: Allocate higher weights to Health Care and Technology during periods of high economic uncertainty
• Crisis Hedging: Reduce exposure to Real Estate and traditional Energy before anticipated market disruptions

For Risk Analysts

• Stress Testing: Use sector-specific volatility patterns from historical crises to model portfolio risk
• Recovery Timelines: Plan liquidity based on observed sector recovery periods (Financials: 3+ years; Technology: <1 year)

For Economic Policymakers

• Targeted Stimulus: Prioritize support for Real Estate and Financials during financial crises; support travel, hospitality, and retail during pandemic-type events
• Industry Monitoring: Focus regulatory attention on sectors showing abnormal volatility patterns

---

🎓 Academic Context

Course: DATS 6101 - Introduction to Data Science
Institution: The George Washington University
Team: Phanindra Kumar Kalaga, Prudhvi Chekuri, Bharat Khandelwal, Dinesh Chandra Gaddam

This project demonstrates:

• Integration of classical statistics with modern machine learning
• Reproducible research practices using R Markdown
• Real-world application of data science to financial markets
• Rigorous hypothesis testing and model validation

---

👥 Authors & Contact

Name	Email
Phanindra Kumar Kalaga	[email protected]
Prudhvi Chekuri	[email protected]
Bharat Khandelwal	[email protected]
Dinesh Chandra Gaddam	[email protected]

---

📖 Documentation

• 📥 INSTALLATION.md - Setup and installation guide
• 📘 Main.html - Complete analysis report (knit analysis/Main.Rmd to generate)
• 📄 Final-Project-Proposal.pdf - Original project proposal

---

🔗 Resources

• GitHub Repository: DATS6101-TeamNeo/final-project
• Dataset Source: S&P 500 GICS on Kaggle
• Python Implementation: Kaggle Notebook

---

📜 License

MIT License - see LICENSE file for details

Copyright (c) 2024 Team Neo

---

🙏 Acknowledgments

• Professor and TAs of DATS 6101 at The George Washington University
• Yahoo Finance for comprehensive historical data
• R and TensorFlow communities for excellent open-source tools

---

⭐ If this research helped your work, please star this repository!

Last Updated: November 2024