FlightDelayStockAnalysis

This project investigates whether operational inefficiencies (measured by flight delays) affect stock market performance for major U.S. airlines.

It integrates Bureau of Transportation Statistics (BTS) flight delay data with Yahoo Finance stock data, stores the results in a PostgreSQL database, and applies Granger causality testing to evaluate potential predictive relationships.

Goal: Determine if airlines with more delays tend to experience lower short-term stock returns.

---

Data Sources

Source	Description	Example Variables
Bureau of Transportation Statistics (BTS)	Monthly U.S. domestic flight data (CSV files)	FL_DATE, ARR_DELAY, CANCELLED, OP_UNIQUE_CARRIER
Yahoo Finance (yfinance)	Daily stock data for major airlines	Close, Open, High, Low, Volume

---

Tools & Technologies

Category	Tools
Language	Python 3
Data Handling	pandas, numpy
Visualization	seaborn, matplotlib
Statistics	scipy, statsmodels
Finance	yfinance
Database	PostgreSQL, SQLAlchemy, psycopg2

---

Methodology

1. Data Cleaning & Missing Value Analysis

• Combined monthly BTS CSVs into a single dataset.
• Visualized missing data with a heatmap.
• Used a Chi-Square test to assess whether missing delays are random across carriers.
• Filled missing ARR_DELAY values with each carrier’s mean delay.

2. Mapping Airlines to Stock Tickers

• Created a mapping of FAA carrier codes to Yahoo Finance tickers.
• Accounted for subsidiaries (e.g., 9E, MQ, OH mapped to AAL).

3. Database Integration

• Stored the cleaned flight data to a PostgreSQL database named flight_data.

4. Stock Data Retrieval

• Downloaded daily stock price data (July 2024 – July 2025) for all mapped tickers.
• Calculated daily stock returns per airline.

5. Statistical Testing

• Verified stationarity using the Augmented Dickey-Fuller (ADF) test.
• Conducted Granger causality tests (lags 1–5) to determine if flight delays predict stock returns.
• Combined results across airlines using Fisher’s method to summarize significance.

---

Outputs

Missingness Check:

Missingness might depend on carrier (not MCAR)
Missing data (%):  0.013376866954650669

Causality Test Results (Per Lag):

1  : 0.5712730644683146 SD:  0.2599717780664828
2  : 0.2200162647400005 SD:  0.3625967528452109
3  : 0.024627090899325303 SD:  0.3140350983941467
4  : 0.11168598652924376 SD:  0.3165748385485352
5  : 0.04019846768413254 SD:  0.27296217582133736

Interpretation: Low p-values (< 0.05) suggest that increases in flight delays may precede declines in airline stock prices in the short term (3 - 5 days). This means that airline operational data can offer some short-term predictive insights into financial performance.

---

How to Run

1. Install dependencies

   pip install pandas numpy matplotlib seaborn scipy statsmodels yfinance sqlalchemy psycopg2

2. Start PostgreSQL and create a database

   CREATE DATABASE flight_data;

3. Update database credentials in the script

   engine = create_engine("postgresql+psycopg2://USERNAME:PASSWORD@localhost:5432/flight_data")

4. Run the analysis

   python analyze_flight_delays.py

---

6. Data Visualization (Tableau Dashboard)

To complement the Python analysis, an interactive Tableau dashboard was created to visualize the key findings.

Dashboard Components:

Sheet 1: Average Flight Delay per Airline

Sheet 2: Stock Returns Comparison

Sheet 3: Correlation Between Mean Delay and Stock Performance (dual-axis bar chart)

These visualizations make it easy to identify which airlines maintain consistent operational efficiency and whether that aligns with their stock performance.

View the Dashboard: View on Tableau Public