HAR-RV IPO Volatility Forecasting Project

Log-Linear Realized GARCH Model for IPO Volatility Prediction Using Peer Group Analysis

Overview

This project implements a sophisticated HAR-RV (Heterogeneous Autoregressive Realized Volatility) forecasting model specifically designed for predicting IPO (Initial Public Offering) volatility using peer group analysis and WRDS Cloud data integration.

Key Features

• WRDS Cloud Integration: Direct connection to WRDS database for real-time CRSP data
• Peer Mapping Algorithm: Intelligent peer selection based on sector and market cap
• Rolling Window Forecasting: Walk-forward validation with HAR-RV model
• Performance Evaluation: Comprehensive metrics vs naive benchmark
• Sample Data Fallback: Robust testing when WRDS is unavailable

Project Structure

HAR_RV_Project/
├── main.py                 # Main pipeline orchestration
├── data_loading.py         # WRDS integration and data processing
├── forecasting.py          # HAR-RV model implementation
├── evaluation.py           # Performance metrics and benchmarks
├── test_wrds.py           # WRDS connectivity testing
├── requirements.txt       # Python dependencies
├── .gitignore            # Git exclusions
└── README.md             # This file

Methodology

1. Data Pipeline

• Loads IPO universe from WRDS Compustat data
• Maps target IPOs to peer groups based on sector and size
• Fetches realized volatility from CRSP daily stock data
• Calculates peer average volatility as fundamental predictor

2. HAR-RV Model

RV_t+1 = β₀ + β₁·RV_t + β₂·Peer_Prior_t + ε_t+1

• Target RV: Absolute daily returns of IPO stock
• Peer Prior: Lagged average volatility of peer group
• Rolling Window: 6-period training window for real-time simulation

3. Evaluation Framework

• RMSE: Root Mean Squared Error
• Naive Benchmark: Random walk forecast
• Skill Score: Percentage improvement over naive

Installation

1. Clone the repository:

git clone https://github.com/ksapru/HAR_RV_Project.git
cd HAR_RV_Project

2. Install dependencies:

pip install -r requirements.txt

3. Set up WRDS credentials:

• Ensure you have WRDS access
• The script will prompt for username and password

Usage

Basic Execution

python3 main.py

Configuration

Edit main.py to customize:

• Sector code (default: 45 for Software)
• Date range (default: 2024)
• Window size (default: 6 periods)
• WRDS username

Sample Data Mode

If WRDS is unavailable, the pipeline automatically falls back to realistic sample data for testing.

Example Output

Loading and preparing data...
Found 47 target IPOs for forecasting.
Building peer mapping...
Target: VHAI | Peers: ADCT.1, ADI, CNLG, MIKR, MANA., PLAB, XETA, DELL, VSTI, CSCO

Loading realized volatility data from WRDS Cloud...
Successfully loaded 138 observations for VHAI

Running HAR-RV forecasting...
Evaluating forecast performance...
Model RMSE: 0.251
Naive RMSE: 0.104
Skill Score (Improvement): -141.67%

Performance Results

Current implementation shows:

• Data Coverage: 138 daily observations for target IPO
• Peer Success: 5/10 peers with valid CRSP data
• Model Performance: Baseline HAR-RV implementation
• Improvement Opportunities: Feature engineering, parameter tuning

Technical Architecture

Data Sources

• WRDS Compustat: IPO universe and fundamentals
• CRSP Daily Stock: Price data for volatility calculation
• Stock Names Database: Ticker to permno mapping

Model Components

• Peer Selection: Sector + market cap filtering
• Volatility Proxy: Absolute daily returns
• Rolling Forecast: Walk-forward validation
• Benchmark: Random walk with drift

Dependencies

• pandas>=1.5.0 - Data manipulation
• numpy>=1.21.0 - Numerical computing
• statsmodels>=0.13.0 - Econometric modeling
• scikit-learn>=1.1.0 - Machine learning utilities
• wrds>=3.4.0 - WRDS Cloud connection

Future Enhancements

1. Feature Engineering

   • Volume-based volatility measures
   • Intraday high-low range
   • Volatility clustering indicators

2. Model Improvements

   • GARCH family extensions
   • Machine learning hybrids
   • Multi-horizon forecasting

3. Data Expansion

   • Cross-asset peer groups
   • Macroeconomic factors
   • Sentiment indicators

4. Production Features

   • Real-time data pipelines
   • Automated model selection
   • Performance monitoring

Research Applications

This framework is suitable for:

• Quantitative Finance: Volatility trading strategies
• Academic Research: IPO market microstructure
• Risk Management: Post-IPO volatility forecasting
• Portfolio Construction: Peer-based factor models

Contributing

1. Fork the repository
2. Create feature branch (git checkout -b feature/AmazingFeature)
3. Commit changes (git commit -m 'Add AmazingFeature')
4. Push to branch (git push origin feature/AmazingFeature)
5. Open Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• WRDS for providing comprehensive financial data
• CRSP for high-quality stock price data
• Compustat for fundamental company information

Contact

Krish Sapru - GitHub

---

Note: This implementation is for research and educational purposes. Production usage requires additional validation, risk management, and compliance considerations.