🚕 NYC Taxi Hidden Congestion Tax Analysis

Winner - HAVK Data Visualization Competition

"The data reveals a problem we're already paying to solve. We're just not solving it."

TL;DR: Rush hour commuters in NYC pay a hidden $3,000/year "congestion tax" through taxi meters running during gridlock. This analysis proves it's real, quantifies the impact, and proposes solutions.

---

📊 The Question

When do NYC taxi passengers pay the most per mile, and why does it matter?

---

🎯 Key Findings

Finding 1: The Rush Hour Premium

Rush hour passengers pay $1.20/mile more (27% premium) than off-peak riders

• Peak cost: $5.62/mile during rush hour (7-10am, 4-7pm weekdays)
• Off-peak cost: $4.42/mile at night (midnight-5am)
• Premium: $1.20/mile difference

---

Finding 2: It's Caused by Traffic Congestion

Speed drops to 8 mph during rush hour (vs 18 mph at night) - inverse correlation r = -0.97

NYC taxi meters charge BOTH:

• Distance: $2.50 per mile when moving
• Time: $0.50 per minute when slow (<12 mph)

Rush hour passengers literally pay for every minute stuck in gridlock.

---

Finding 3: It's a Commuter Tax

Weekday rush hours cost $0.73/mile MORE than weekend same hours

• Weekday rush hour: $5.39/mile at 8 mph
• Weekend same hours: $4.66/mile at 16 mph
• Commuter penalty: 15.7% extra for having a 9-to-5 job

---

💰 Economic Impact

$1.20    extra per mile during rush hour
$3,000   annual cost for typical commuter (5-mile commute × 2 trips/day × 250 workdays)
$500M+   estimated citywide impact per year (conservative estimate)

The Problem

This "hidden congestion tax" is:

• INVISIBLE - passengers don't see it until the ride ends
• REGRESSIVE - punishes workers who must commute during peak hours
• WASTED - money compensates drivers for lost time but doesn't reduce congestion

---

🔬 Methodology

Dataset

• Source: NYC Taxi & Limousine Commission (TLC)
• Time Period: March 2016
• Total Records: 12,210,952 trips
• Sample Used: 100,000 trips (1% stratified random sample)

Key Engineered Features

Feature	Calculation	Purpose
hour	Extracted from pickup_datetime	Temporal pattern analysis
trip_duration_min	(dropoff - pickup) / 60	Speed calculation
speed_mph	(distance / duration) × 60	Congestion proxy
fare_per_mile	fare_amount / trip_distance	Primary metric
is_weekend	day_of_week in [5,6]	Commuter isolation

Analysis Pipeline

Raw Data (12M rows)
    ↓
Random Sampling (100K rows, seed=42)
    ↓
Feature Engineering (time, speed, cost efficiency)
    ↓
Data Cleaning (outlier removal, geographic filtering)
    ↓
Exploratory Analysis (temporal patterns)
    ↓
Hypothesis Testing (congestion causation)
    ↓
Validation (weekday/weekend comparison)
    ↓
Policy Recommendations

Data Cleaning

Removed outliers and impossible values:

• Trip distance: 0-100 miles
• Fare amount: $2.50-$250 (NYC minimum to reasonable max)
• Passenger count: 1-6 (legal limit)
• Trip duration: 1-120 minutes
• Speed: 2-60 mph (realistic city speeds)
• Geographic bounds: NYC metropolitan area

Retention rate: 97.4% (96,991 trips retained)

---

📈 Visualizations

1. The Rush Hour Premium

Shows when passengers pay the most per mile throughout the day.

Key Insight: Business hours (6am-7pm) show consistent premium over late night/early morning.

2. The Congestion Proof

Dual-axis chart showing inverse correlation between speed and cost per mile.

Key Insight: When traffic slows to 8 mph during rush hour, cost spikes to $6/mile. Correlation r = -0.97 (p < 0.001).

3. The Commuter Penalty

Weekday vs weekend comparison isolating the commuter effect.

Key Insight: Weekday rush hours cost significantly more than weekend same hours - proof this is about work commutes, not general congestion.

---

💡 Policy Recommendations

Immediate Actions (0-2 years)

1. Transparent Receipts

• Mandate receipts show breakdown: distance charge vs time-in-traffic charge
• Creates awareness without new regulations
• Who: NYC Taxi & Limousine Commission (TLC)
• Cost: Minimal (software update)

2. Peak/Off-Peak Pricing

• Implement time-based meter rates (like electricity or transit)
• Peak: $3.00/mile + $0.60/min | Off-peak: $2.25/mile + $0.40/min
• Revenue-neutral for drivers
• Who: TLC with driver union approval
• Cost: Meter recalibration + education campaign

3. Congestion Relief Surcharge

• Add $0.50 per rush-hour trip → dedicated fund
• Revenue (~$20M/year) funds: bus lanes, smart signals, routing systems
• Who: NYC Council + TLC
• Cost: Revenue-positive

Medium-Term (2-5 years)

4. Taxi Express Lanes

• Designate 1 lane on major routes for taxi/bus only during rush hours
• Who: NYC DOT
• Cost: $50-100M (infrastructure)

5. Smart Routing Systems

• Mandate real-time routing for all taxis
• Who: TLC + tech vendors
• Cost: $20-30M (software)

---

🛠️ Technical Stack

# Core libraries
import pandas as pd              # Data manipulation
import numpy as np               # Numerical operations
import matplotlib.pyplot as plt  # Visualization
import seaborn as sns           # Statistical visualization
from scipy.stats import pearsonr # Correlation testing

# Key techniques
- Stratified random sampling (preserves time distribution)
- Feature engineering (temporal, speed, efficiency metrics)
- Statistical hypothesis testing (Pearson correlation)
- Outlier detection and removal (IQR method)
- Geospatial filtering (bounding box)

---

📁 Repository Structure

nyc-taxi-hidden-congestion-tax/
├── yellow_tripdata_2016-03.csv    # Original dataset (not included - 2GB, download from source)
├── notebook.ipynb                  # Main analysis notebook
├── images/
│   ├── cost-per-mile-per-hour.png              # Visualization 1
│   ├── speed-vs-cost.png              # Visualization 2
│   └── weekday-vs-weekend.png            # Visualization 3
├── presentation.pdf                      # Competition presentation
├── README.md                           # This file
├── requirements.txt                    # Python dependencies
└── LICENSE                             # MIT License

---

🚀 Quick Start

Prerequisites

Python 3.8+
pip

Installation

1. Clone the repository

git clone https://github.com/dganesh06/Analyzing-Traffic-Congestion-and-Effect-on-Taxi-Prices-in-NYC.git
cd Analyzing-Traffic-Congestion-and-Effect-on-Taxi-Prices-in-NYC

2. Install dependencies

pip install -r requirements.txt

3. Download the dataset

• Visit NYC TLC Trip Record Data
• Download "Yellow Taxi Trip Records - March 2016"
• Place in data/ directory as yellow_tripdata_2016-03.csv

4. Run the analysis

jupyter notebook notebooks/analysis.ipynb

---

📊 Reproducing Results

The analysis uses a seeded random sample for reproducibility:

np.random.seed(42)  # Set seed for reproducibility
n_sample = 100000   # Target 100K rows from 12.2M

Running the notebook with this seed will produce identical results to the competition submission.

Expected runtime: 5-10 minutes on a standard laptop

---

🎓 Key Learnings

What Worked

✅ Started with a human question - "When do passengers get ripped off?" vs "What patterns exist in taxi data?"

✅ Proved causation, not just correlation - Speed data confirmed congestion hypothesis

✅ Validated with controls - Weekday/weekend split ruled out alternative explanations

✅ Made it actionable - Specific policy recommendations with cost estimates

Technical Highlights

• Smart sampling: 1% sample preserved patterns, enabled rapid iteration
• Feature engineering: fare_per_mile was the killer metric - simple, intuitive, powerful
• Dual-axis visualization: Inverted speed axis created intuitive "both getting worse" visual
• Statistical rigor: Correlation testing (r = -0.97, p < 0.001) made findings bulletproof

---

🏆 Competition Results

HAVK Data Visualization Competition - 1st Place

Judge Feedback:

"This analysis didn't just find a pattern - it went beyond by connecting it to a real-world scenario and proposing insights and solutions from the visualization."

---

🔗 Related Work

Context: NYC Congestion Pricing (2025)

In January 2025, NYC implemented congestion pricing - charging cars $9-15 to enter Manhattan below 60th St during peak hours. This analysis revealed that taxi passengers were already paying a hidden congestion tax long before the official policy.

Follow-up Questions

• How did 2025 congestion pricing affect taxi passenger costs?
• Do Uber/Lyft show similar patterns?
• How does NYC compare to London, Singapore, or Tokyo?
• Can machine learning predict optimal routing to minimize passenger costs?

---

📚 References

1. NYC Taxi & Limousine Commission - Trip Record Data
2. NYC Taxi Fare Structure - Official Rates
3. NYC Congestion Pricing - MTA Central Business District Tolling Program
4. Urban Transportation Economics - Vickrey, W. (1969) "Congestion Theory and Transport Investment"

---

📬 Contact

Author: [Divya Ganesh]

• 📧 Email: [email protected]
• 💼 LinkedIn: linkedin.com/in/divyaganesh05
• 🐙 GitHub: @dganesh05

---

📄 License

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

---

🙏 Acknowledgments

• HAVK for organizing the data visualization competition
• NYC TLC for making taxi data publicly available
• Urban planning community for inspiration on transportation equity issues
• Competition judges for valuable feedback

---

⭐ If You Found This Useful

• Star this repository
• Share on LinkedIn/Twitter
• Use the methodology for your own city's transportation data
• Contribute improvements via pull request

---

🚧 Future Work

Potential Extensions:

• Compare to Uber/Lyft pricing patterns
• Analyze impact of 2025 congestion pricing on taxi usage
• Build interactive dashboard (Plotly/Streamlit)
• Machine learning model to predict optimal routes
• Cross-city comparison (NYC vs SF vs Chicago)
• Time series analysis (2016 vs 2020 vs 2025)

Contributions welcome!

---

Built with 🚕 by someone who cares about urban transportation equity

⬆ Back to Top