81 lines
4.7 KiB
Markdown
81 lines
4.7 KiB
Markdown
# Texas RRC Inspection Expenses Analysis
|
||
|
||
Empirical analysis of whether organizational capacity (budget and staffing) predicts regulatory outputs at the Texas Railroad Commission (RRC), using a fixed-effects panel model across 13 districts from FY2016–2025.
|
||
|
||
## Research Questions
|
||
|
||
Does OGI budget and FTE predict more inspections, higher compliance rates, and faster violation resolution? How is that relationship moderated by goal ambiguity, district-level heterogeneity, and geographic context?
|
||
|
||
**Four hypotheses:**
|
||
- **H1 — Capacity → Outputs:** Higher OGI budget and FTE predict better regulatory performance
|
||
- **H2 — Goal Ambiguity:** Higher share of non-inspection budget weakens the capacity-output relationship
|
||
- **H3 — District Heterogeneity:** The capacity-output relationship varies across RRC districts
|
||
- **H4 — Geographic Moderation:** Offshore and border-proximate districts moderate the capacity-output relationship
|
||
|
||
## Data
|
||
|
||
| Source | Description |
|
||
|--------|-------------|
|
||
| RRC Administrative Database (PostgreSQL) | ~1.9M inspection events, ~193K enforcement actions, FY2016–2025, 13 districts |
|
||
| `RRC Budget Data.xlsx` | Legislative Appropriations Requests, FY2016–2024, disaggregated by goal/strategy |
|
||
|
||
**Unit of analysis:** District-year (13 districts × 10 years; regression sample N = 104, FY2016–2023)
|
||
|
||
**Key constraint:** Budget varies only at the statewide level; identification relies on year-to-year temporal variation within districts.
|
||
|
||
## Measures
|
||
|
||
**Dependent variables:**
|
||
- Total inspections per district-year
|
||
- Compliance rate (% of inspections with no failure)
|
||
- Violation resolution rate (% resulting in operator compliance on re-inspection)
|
||
- Average days-to-enforcement
|
||
|
||
**Key independent variables:**
|
||
- OGI total appropriations ($M) — primary capacity measure
|
||
- OGI authorized FTE — auxiliary capacity measure
|
||
- Inspection budget share = OGI / (OGI + ERD) — goal ambiguity proxy (range: 0.59–0.67)
|
||
- Offshore district (Districts 02, 03, 04) and border district (Districts 01–04) indicators
|
||
|
||
## Methods
|
||
|
||
Fixed-effects panel regression with district fixed effects and district-clustered standard errors. Four model specifications for H1–H4. Robustness checks include wild cluster bootstrap (B = 999, Rademacher weights) for small-sample inference (G = 13 clusters) and a one-year distributed lag specification.
|
||
|
||
## Key Findings
|
||
|
||
**H1 — Supported with caveats:** Each $1M increase in OGI budget is associated with ~666 additional inspections (p < .01), +0.26 pp compliance rate (p = .02), and +1.05 pp resolution rate (p < .01). *However*, wild cluster bootstrap p-values (≈0.49–0.51) indicate asymptotic inference substantially overstates precision; results are directionally consistent but do not survive bootstrap testing with only 13 clusters.
|
||
|
||
**H2 — Exploratory:** Budget × inspection-share interaction is negative and significant for compliance (β = −6.53, p < .01), consistent with diminishing returns as the inspection mandate becomes better-resourced, but cannot rule out year-specific confounders given budget share varies only over time.
|
||
|
||
**H3 — Descriptive:** Budget-compliance slopes range from −0.34 pp/$1M (District 03) to +1.36 pp/$1M (District 6E). Near-perfect multicollinearity renders standard errors unreliable; slopes are descriptive only.
|
||
|
||
**H4 — Partial support:** Offshore districts show +7.6 pp baseline compliance (p = .02); border districts show +6.0 pp (p = .03). These are level effects—budget sensitivity does not significantly differ by geography. Moran's I = −0.051 (no spatial autocorrelation).
|
||
|
||
**Descriptive trends (2016–2023):** OGI budget +86%, mean inspections/district +85%, mean compliance rate +8.5 pp (83.1% → 91.6%), mean resolution rate +32.9 pp (36.8% → 69.7%).
|
||
|
||
## Limitations
|
||
|
||
1. Budget identified from temporal variation only (statewide, not district-level allocation)
|
||
2. Small panel: N = 104 over 8 years limits statistical power and slope heterogeneity estimation
|
||
3. Wild cluster bootstrap undercuts asymptotic inference; results should be treated as suggestive
|
||
4. Parallel pre-trends assumption cannot be directly tested
|
||
5. District-level budget allocation unobserved; some decentralization unmeasured
|
||
6. FY2025 enforcement metrics subject to right-censoring
|
||
|
||
## Repository Structure
|
||
|
||
```
|
||
texas_inspection_expenses.ipynb # Main analysis notebook
|
||
RRC Budget Data.xlsx # Budget source data
|
||
requirements.txt # Python dependencies
|
||
texas_inspection_expenses.html # Rendered notebook export
|
||
```
|
||
|
||
## Setup
|
||
|
||
```bash
|
||
pip install -r requirements.txt
|
||
```
|
||
|
||
Requires a PostgreSQL connection to the `texas_data` database with RRC inspection/violation tables. Configure credentials via environment variables (see `python-dotenv`).
|