Do data centers lift local economies?
The evidence is still emerging.

We applied propensity-matched 2×2 difference-in-differences estimation to US counties hosting new hyperscale and large-enterprise data centers, comparing each against up to five matched counties that did not receive a facility during the same period.

County identifiers are anonymized in the summary. The full dataset with FIPS codes, bootstrapped CIs, and suppression fractions is available for download after registration.

What We Measured

• DC-adjacent sector employment (BLS QCEW, 7-NAICS bundle: construction, data infrastructure, IT services, utilities)
• County sales-tax revenue (Texas Comptroller — Texas counties only)
• Hotel occupancy tax receipts (Texas Comptroller — Texas counties only)

Estimator

Propensity scores estimated via logistic regression on five pre-treatment covariates (sector employment CAGR, industry mix index, sales-tax CAGR, HOT CAGR, employment level). Caliper matching, top 5 controls per treated county. Canonical 2×2 DiD. Bootstrapped 95% confidence intervals, 200 replications.

QCEW windows: ±8 quarters around the treatment quarter. Fiscal windows: ±24 months around the treatment month. Null/missing observation slots counted as suppressed and excluded from estimation.