China’s national AI pilot zones have lifted local entrepreneurship: designation is associated with higher new firm formation per capita, especially in cities with weak factor markets and strong transport links, and the policy generates spillovers within provinces.

Title: Has the artificial intelligence innovation and development pilot zone improved regional entrepreneurial activity? Author: Jaiyi Song (The University of Hong Kong) Journal: Journal of Applied Economics and Policy Studies, Vol.19 Issue 9 (Available Online 24 Aug 2026) DOI: 10.54254/2977-5701/2026.36306

Main Finding

The designation of National New‑Generation Artificial Intelligence Innovation and Development Pilot Zones significantly increases regional entrepreneurial activity (measured as registered enterprises per 10,000 residents). The effect operates through three main channels — stimulating the digital economy, promoting offline experiential consumption/economies, and improving regional R&D efficiency — and shows important spatial and city‑level heterogeneity. The policy also produces positive spillovers to other cities within the same province.

Key Points

• Causal design: The pilot‑zone designation is treated as a quasi‑natural, staggered policy shock; a multi‑period difference‑in‑differences (DID) framework is used to identify effects.
• Measured outcome: Urban entrepreneurial activity proxied by number of registered enterprises per 10,000 residents.
• Mechanisms identified:
  • Digital economy vitality: pilot zones accelerate digital adoption and e‑commerce ecosystems, expanding market demand and low‑barrier entrepreneurial niches.
  • Offline experiential economy: development of offline scenario/application spaces increases demand and business models that complement digital services.
  • R&D efficiency: concentrated AI resources and policy supports raise local R&D efficiency, reducing supply‑side costs for new ventures.
• Political and ecosystem channels emphasized: (a) government incentive effect (fiscal/tax incentives, regulatory flexibility and infrastructure attracted by pilot status) and (b) entrepreneurial climate spillover (demonstration effects, labor mobility, knowledge diffusion from AI firms).
• Heterogeneity:
  • Stronger policy effects in cities with relatively underdeveloped factor markets but good transportation infrastructure.
  • Positive correlates at city level: higher per‑capita GDP, greater internet penetration, higher education levels.
  • Negative/limiting correlates: excessive concentration of financial and fiscal resources (which can inhibit new firm entry).
• Spatial effects: evidence of intra‑provincial spillovers (the paper examines leading, siphoning, and diffusion patterns), indicating benefits beyond the treated city.
• Research gap addressed: distinguishes localized policy effects of pilot zones from nationwide, platform‑wide AI diffusion as a General Purpose Technology.

Data & Methods

• Data: Panel data for 285 prefecture‑level (and above) Chinese cities, covering 2009–2023.
• Treatment: designation as a National New‑Generation Artificial Intelligence Innovation and Development Pilot Zone (staggered adoption across cities).
• Identification strategy: multi‑period/staggered difference‑in‑differences (DID) comparing treated and control cities before and after designation; controls included to net out nationwide AI diffusion and other confounders.
• Outcome variable: registered enterprises per 10,000 residents (narrow measure of entrepreneurial vitality).
• Additional analyses: mechanism tests (digital economy indicators, offline experiential economy measures, R&D efficiency), heterogeneity tests by city characteristics (factor market development, transport infrastructure, Internet, education, fiscal/financial concentration), and spatial spillover/interaction analyses across cities within provinces.
• Robustness: the paper reports controls and tests to separate national GPT effects from place‑based policy effects and explores spatial patterns (details on placebo and parallel‑trend checks are described in the full paper).

Implications for AI Economics

• Place‑based AI policy matters: Beyond the nationwide diffusion of AI as a GPT, targeted pilot‑zone policies generate incremental, place‑specific gains in entrepreneurship via resource concentration, policy incentives, and local demonstration effects.
• Complementarity of demand and supply channels: Effective AI policy should target both supply (R&D, human capital, innovation resources) and demand (digital ecosystems, offline application scenarios) to unlock entrepreneurial opportunities broadly across sectors.
• Heterogeneous targeting: Policies will be relatively more effective in cities with weaker factor markets but adequate physical connectivity; one‑size‑fits‑all national approaches risk underperforming in different local contexts.
• Caution on resource concentration: Overconcentration of finance or fiscal resources in a few institutions/activities can inhibit broad‑based entrepreneurship; policy should promote inclusive access to finance and avoid excessive centralization of fiscal supports.
• Spatial coordination: Because pilot zones create intra‑provincial spillovers, provincial‑level coordination (rather than purely city‑level competition) can magnify benefits and reduce negative siphoning effects.
• Policy recommendations (high level): reinforce digital infrastructure and Internet penetration, invest in education and R&D efficiency, develop offline scenario/application spaces for AI, design fiscal/financial incentives to be inclusive, and coordinate policy across neighboring jurisdictions to maximize positive spillovers.

If you want, I can extract the paper's key tables/figures, summarize the identification/robustness diagnostics in more detail, or produce a slide‑ready 1‑page summary for policy audiences.