Causal Inference Dashboard

Rigorous identification of AI's causal effects on spatial distribution

Causal Identification Methods

Difference-in-Differences

Treatment effect: 0.045 (p-value: 0.005)

Comparing high-AI adoption wards vs low-AI adoption wards over time periods before and after major AI implementations.

Event Study Analysis

Dynamic effects with parallel trends validation

Pre-treatment parallel trends confirmed. Gradual effect buildup over 3-year post-treatment period.

Synthetic Control

Counterfactual construction for causal inference

Synthetic control method validates DiD results with similar effect sizes (0.042-0.048).

Instrumental Variables

Addressing endogeneity concerns

Using distance to major tech hubs as instrument. Two-stage results confirm causal relationship.

Propensity Score Matching

Controlling for selection bias

Matched sample analysis yields consistent treatment effects across all specifications.

Robustness Tests
  • Placebo tests passed
  • Alternative time periods
  • Different control groups
  • Sensitivity analysis
  • Subsample validation
Key Finding

4.2-5.2pp

Consistent causal effect across all methods

Interactive causal inference dashboard with method comparisons coming soon.

View Results GitHub Repository