Autocracy and the Innovation Frontier: Can Closed Regimes Produce Technology?

The autocracy ceiling: what Russia lost and what China can't yet buy
Russia is the failure mode. China is testing whether scale alone can break it.

The growth-rate gap between democracies and autocracies, Pritchett (2012) found, averaged just 0.2 percentage points over 1980–2008. Variance was the bigger story: 2.5× higher for autocracies. Albornoz et al. ran the question on science output and found authoritarianism leaves physics, math and chemistry alone. Catch-up isn't the problem. The frontier is.

Russia spent the last decade rebuilding itself as the case study for what doesn't work. R&D has been stuck near 1% of GDP, about a third of the OECD average, and the headcount of researchers fell again in 2023. Novaya Gazeta tracked 2,500 scientists out by counting affiliation changes in publications; the IT industry association told parliament 70,000 IT workers had left by March 2022. Web of Science stopped accepting Russian journal applications, and the EU's REG 2024/1745 cut Russians out of EU patent offices entirely. On April 30 Bloomberg reported that 90%+ of Russia's sanctioned tech now comes through China. The official "import substitution" program is a Chinese supply chain.

China runs the harder experiment. R&D in 2025 reached $569B at 2.8% of GDP, close to the OECD average and well above the EU's 2.11%. Basic-research funding crossed 7% of total R&D. DeepSeek, Kimi and MiniMax overtook US models in OpenRouter API calls in February. Nvidia's CEO says half the world's AI researchers are Chinese. BYD outsold Tesla in Europe. By 2020 China had passed the US in clinical-trial volume.

Two limits haven't moved. SMEE makes 28nm DUV lithography, and CSIS doesn't expect domestic sub-10nm production before 2030. By Tyler Cowen's count, the ten most-influential science papers of the last decade were all American. And in March 2026 Beijing barred Manus AI's CEO and chief scientist from leaving the country during the review of their $2B Meta acquisition — exactly what happened to Jack Ma in 2020.

Acemoglu and Robinson called this in 2010: politically connected incumbents kill creative destruction because new firms displace them. Bardhan put a number on it — China had fewer corporate insolvencies than Romania in 2017. North added the path-dependence story: Russia's tech-trade balance fell from +$20M in 2000 to −$1B in 2009 (Klochikhin). Iran shows the same shape in extreme form, with the IRGC owning 35–50% of GDP, exempt from taxation and audits.

Two questions decide the next five years. Will domestic Chinese lithography reach below 10nm? Will any Chinese-origin paper crack the global top decile of influence? If neither happens despite 2.8% of GDP and half the world's AI researchers being Chinese, the ceiling holds. If one does, the framework needs rewriting.