The problem

The CHIPS Act bets tens of billions on broadening where American semiconductor innovation happens. But whether federal R&D dollars actually move the geography of invention — rather than reinforcing incumbent hubs — is an open empirical question, and the county level is where the policy mechanisms (labor markets, knowledge diffusion, supplier networks) would have to show up.

Technical approach

A newly constructed county-year panel links CPC-filtered USPTO patenting (PatentsView, with federal assignees screened out by type-code and keyword filters) to DoD microelectronics R&D obligations from USAspending, with BEA/BLS/NSF controls. Two-way fixed-effects models estimate lagged direct effects; distance-band models test whether proximity to funded counties raises non-recipient patenting.

Technical highlights

• Direct effects are short-run and specification-sensitive: positive at a one-year horizon, dissipating after — federal R&D activates existing capability rather than building new regional capacity.
• The headline null: once direct funding is accounted for, physical proximity confers nothing. The "triangular alliance" geography of the Cold War has become vertical, institutionally-gated silos.
• Policy teeth: as currently structured, defense contracting is unlikely to achieve the CHIPS Act’s spatial ambitions — funding intensity, not geographic proximity, drives where innovation happens.

Why it matters here

For patent/IP audiences: demonstrated fluency with USPTO patent data at scale — CPC taxonomy, assignee disambiguation, inventor geography — applied to a live question in innovation policy. This is the analytical foundation of my patent-track direction.