AI Infrastructure and Regional Electricity Demand: Evidence from U.S. Interconnection Queues

This package contains the replication materials for Lamb (2026), \"AI Infrastructure and Regional Electricity Demand: Evidence from U.S. Interconnection Queues.\" The paper studies whether the U.S. data center buildout has produced a measurable signature in regional electricity demand, using a four-layer identification strategy (panel regression with wild cluster bootstrap, synthetic control, difference-in-differences, and narrative validation) and a three-model forecasting pipeline (ARIMA, Prophet, XGBoost). The main empirical result is a 34.8 index-point gap in ERCOT minimum hourly demand against a synthetic counterfactual constructed from low-exposure balancing authorities. The package includes the analysis-ready panels, all Python and Stata scripts, final figures and tables, and the working paper PDF. Raw third-party data is not redistributed; the README documents how to acquire it. Code is released under the MIT License; derived data and outputs under CC BY 4.0.