# Texas District Analysis: Regulatory Transparency and Enforcement in the Oil & Gas Industry

A research project examining how transparency disclosure reforms affect enforcement behavior in the Texas Railroad Commission (RRC), with a focus on district-level heterogeneity across 13 RRC regulatory districts from 2015–2025.

## Research Overview

**Core question**: Does making well-level violation data publicly searchable change how quickly the RRC acts on violations?

The January 2019 RRC policy change — making well violation data publicly searchable — serves as the exogenous policy shock. The analysis tests whether and how this disclosure reform altered enforcement timing and compliance outcomes across districts, with particular attention to offshore-regulating districts (02, 03, 04) and structural moderators like basin composition, enforcement capacity, and environmental justice dimensions.

**Key findings:**
- No immediate post-2019 level shift in enforcement timing (coef=0.1514, p=0.33)
- Significant post-2019 trend acceleration: enforcement speed improves gradually over time (coef=−0.3603, p=0.001)
- Offshore-regulating districts show differential post-policy response (coef=0.3819, p<0.001), strongest in 2023–2024
- Basin composition is the clearest structural correlate of district-level heterogeneity

## Data

All raw data originates from the Texas Railroad Commission and supplementary government sources:

| Source | Description | Size |
|--------|-------------|------|
| Texas RRC | ~3.6M inspection records (pipe-delimited) | 424 MB |
| Texas RRC | ~368K violation records (pipe-delimited) | 66 MB |
| U.S. Census | Poverty rates and demographics by census tract | — |
| USDA RUCA (2020) | Rural-Urban Commuting Area classifications | 25 MB |
| USEIA | Shale basin and play shapefiles | — |
| Texas county shapefiles | County boundaries for spatial visualization | — |

Data covers approximately 1.01 million wells, 1.87 million inspections, and 191K violations within the 2015–2025 study window.

**Note**: Raw data files are large (several hundred MB each) and are excluded from version control via `.gitignore`. The data pipeline is fully documented in the `rebuild/` notebooks.

## Repository Structure

```
texas-district-analysis/
├── analysis/
│   ├── well_analyzer.py                          # Core analysis engine (PostgreSQL → metrics)
│   ├── updated_district_level_analysis_2015-2025_offshore_controls.ipynb  # Main analysis notebook
│   ├── draft.md                                  # Manuscript draft
│   ├── draft_appendix.md                         # Technical appendix with model specifications
│   ├── *.png                                     # Figures (event study, district maps, etc.)
│   └── archive/                                  # Earlier notebook versions and alternate specs
├── data/
│   ├── INSPECTIONS.txt                           # Raw inspection records (pipe-delimited)
│   ├── VIOLATIONS.txt                            # Raw violation records (pipe-delimited)
│   ├── RUCA-codes-2020-tract.csv                 # RUCA classification by census tract
│   ├── district_by_county.csv                    # District–county crosswalk
│   └── {oil_gas_basin,shale_play,texas_county,texmex}_shape/  # ESRI shapefiles
├── rebuild/
│   ├── rrc_api_data.ipynb                        # Step 1: Fetch and process RRC API data
│   ├── create_violations_inspections.ipynb       # Step 2: Build cleaned inspection/violation files
│   ├── add_census_data.ipynb                     # Step 3: Link census demographics
│   ├── add_shape_layers.ipynb                    # Step 4: Spatial feature engineering
│   ├── well_shape.ipynb                          # Step 5: Well geometry and shapefile creation
│   └── well-api-manual.pdf                       # RRC API technical documentation
├── papers/                                       # Manuscript versions (DOCX + PDF)
├── analysis_output.json                          # Pre-computed summary statistics
└── requirements.txt                              # Python dependencies
```

## Analysis Pipeline

```
Raw RRC Data (API)
    ↓  rebuild/rrc_api_data.ipynb
Cleaned Inspections & Violations CSVs
    ↓  rebuild/create_violations_inspections.ipynb
Link Census Demographics
    ↓  rebuild/add_census_data.ipynb
Add Geographic Layers
    ↓  rebuild/add_shape_layers.ipynb
PostgreSQL Data Warehouse
    ↓  analysis/well_analyzer.py
District-Year Panel
    ↓  analysis/updated_district_level_analysis_2015-2025_offshore_controls.ipynb
Econometric Models → Figures → Manuscript
```

### Econometric Models

| Model | Description |
|-------|-------------|
| 1 | Interrupted time-series (all districts pooled) |
| 2 | District-specific post-policy fixed effects |
| 3 | Offshore jurisdiction moderator (districts 02/03/04) |
| 4 | Spatial autocorrelation diagnostics (Moran's I, 5,000 permutations) |
| 5 | Structural moderators: capacity, baseline compliance, EJ, geology, rurality, border proximity |

## Setup

### Prerequisites

- Python 3.9+
- PostgreSQL (with PostGIS for spatial queries)
- The `well_analyzer.py` module reads database credentials from environment variables

### Install dependencies

```bash
pip install -r requirements.txt
```

### Database configuration

Set the following environment variables before running analysis:

```bash
export PGHOST=localhost
export PGPORT=5432
export PGUSER=your_user
export PGPASSWORD=your_password
export PGDATABASE=texas_data
```

### Run the data pipeline

Execute the notebooks in `rebuild/` in order (steps 1–5) to populate the PostgreSQL database, then open `analysis/updated_district_level_analysis_2015-2025_offshore_controls.ipynb` for the main analysis.

## Key Statistics (2015–2025)

- **1,878,764** inspections across **420,185** unique wells
- Overall compliance rate: **89.9%** (up from 88.4% in 2015 to 92.9% in 2024)
- **193,338** violations across **81,670** unique wells
- Mean days from violation discovery to enforcement action: **127** (median: 14)
- Compliance on re-inspection: **57.2%**
- District compliance range: 81.2% (District 09) to 94.4% (District 8A)

## Dependencies

```
pandas
numpy
sqlalchemy
psycopg2
scipy
statsmodels
matplotlib
seaborn
geopandas
shapely
libpysal
esda
```