# Texas Borderlands: Regulatory Enforcement Disparities

An empirical research project examining whether oil and gas regulatory enforcement in Texas differs systematically between border-proximate and interior districts — and whether a 2019 disclosure reform produced heterogeneous effects across those regions.

## Research Questions

- **RQ1:** Do border-exposed Texas Railroad Commission (RRC) districts differ from non-border districts in inspection intensity, violation detection, enforcement timing, and resolution rates?
- **RQ2:** Did the 2019 disclosure reform change enforcement outcomes differently in border districts versus non-border districts?

## Key Findings

| Outcome | Border Districts | Non-Border Districts |
|--------|-----------------|---------------------|
| Inspections per well | 1.329 | 1.515 |
| Violations per inspection | 0.130 | 0.098 |
| Days to enforcement | 145.2 | 122.8 |
| Resolution rate | 0.543 | 0.596 |

**Post-2019 reform effect:** Enforcement processing time in border districts improved by **~75 days** (p=0.016) relative to non-border districts — but inspection reach and resolution rates did not converge. Conclusion: *faster pipeline, not wider pipeline*.

## Project Structure

```
texas-borderlands/
├── analysis/
│   ├── borderlands.ipynb                   # Main analysis notebook
│   ├── well_analyzer.py                    # WellAnalyzer class (data loading + metrics)
│   └── output_borderlands/
│       ├── rq1_results.csv                 # RQ1 regression results
│       ├── rq2_results.csv                 # RQ2 FE interaction results
│       ├── district_year_panel_borderlands.csv
│       ├── border_vs_nonborder_trends.png
│       ├── money_plot_timing_border_prepost2019.png
│       ├── well_border_exposure_map.png
│       ├── continuous_exposure_results.csv
│       ├── cutoff_sensitivity_results.csv
│       └── border_type_split_results.csv
│
├── data/
│   ├── oil_gas_basin_shape/                # EIA TX shale basin boundaries
│   ├── shale_play_shape/                   # EIA TX shale play delineations
│   ├── texas_county_shape/                 # US Census TX county subdivisions (2025)
│   ├── texmex_shape/                       # US Census TX-MX international boundary (2023)
│   ├── competition_panel.csv
│   └── district_competitor_links.csv
│
├── intro_thoery_methods_analysis_results_discussion.md  # Full paper draft
├── appendix.md                             # Supplementary tables and robustness checks
└── requirements.txt
```

## Tech Stack

- **Python 3**
- **pandas / numpy** — data manipulation and panel construction
- **sqlalchemy / psycopg2** — PostgreSQL database access
- **geopandas / shapely** — geospatial analysis and border proximity measurement
- **scipy / statsmodels** — regression models (OLS, fixed effects)
- **libpysal / esda** — spatial econometrics
- **matplotlib / seaborn** — visualization
- **python-dotenv** — environment configuration

## Setup

### 1. Install dependencies

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

### 2. Configure the database connection

Create a `.env` file in the project root (or set environment variables directly):

```env
PGHOST=localhost
PGPORT=5432
PGUSER=postgres
PGPASSWORD=your_password
PGDATABASE=texas_data
```

The database should have PostGIS enabled and contain the following tables:

| Table | Description |
|-------|-------------|
| `well_shape_tract` (or similar) | Wells with location and demographic enrichment |
| `inspections` | Inspection records with dates and district info |
| `violations` | Violation records with enforcement and resolution dates |

The `WellAnalyzer` class auto-detects the wells table name from a set of known aliases.

### 3. Run the analysis

Open and run the Jupyter notebook:

```bash
jupyter notebook analysis/borderlands.ipynb
```

Or use the `WellAnalyzer` class directly:

```python
from analysis.well_analyzer import WellAnalyzer

analyzer = WellAnalyzer()
analyzer.print_analysis()
analyzer.export_analysis("output.json")
```

## Data

### Primary Data (PostgreSQL)

- **~1.01M wells** with geospatial coordinates and demographic/census tract enrichment
- **~1.87M inspections** (2015–2025)
- **~191.7K violations** (2015–2025)
- **District-year panel:** 143 observations (13 RRC districts × 11 years)

### Shapefiles

| File | Source | Purpose |
|------|--------|---------|
| `texmex_shape/` | US Census Bureau (2023) | TX-MX border geometry for proximity calculations |
| `texas_county_shape/` | US Census Bureau (2025) | State and county boundaries |
| `oil_gas_basin_shape/` | US EIA | Texas shale basin delineations |
| `shale_play_shape/` | US EIA | Texas shale play delineations |

### Border Exposure Definitions

- **District-level:** Binary — district centroid or wells within 50 km of any state/international border
- **Well-level:** Binary flags at 25 km and 50 km buffers from TX-Mexico border
- **Border subtypes:** TX-MX, TX-NM, TX-OK, TX-LA

Border-exposed wells (50 km buffer): **169,520** of 1,010,432 total.

## Empirical Design

**Unit of analysis:** Texas RRC district × year (2015–2025)

**Outcome variables:**
- Inspection intensity (inspections per well)
- Violation rate (violations per inspection)
- Days to enforcement action
- Resolution rate (compliance on reinspection)

**RQ1 — Levels model:**
```
Y_{dt} = α + β·Border_d + γ·X_{dt} + ε_{dt}
```

**RQ2 — Fixed effects interaction model:**
```
Y_{dt} = α_d + δ_t + β·(Post2019_t × Border_d) + γ·X_{dt} + ε_{dt}
```

**Robustness checks:** Border-type splitting, continuous exposure shares, cutoff sensitivity (25/75/100 km thresholds).

## Documentation

- `intro_thoery_methods_analysis_results_discussion.md` — full paper draft covering theory, methods, results, and discussion
- `appendix.md` — supplementary regression tables, robustness checks, and district profiles

## License

This project is for academic research purposes. Underlying data sources are public records from the Texas Railroad Commission and US federal agencies.