An Evidence-Based Account of the ARPA Model in Europe

Jasper Doeninghaus, Harvard Kennedy School, 2026. SPRIND, Germany's Federal Agency for Disruptive Innovation, faces two strategic challenges that share an evidentiary foundation but address distinct audiences. The first is domestic legitimation. The German federal government must decide within the current legislative period whether to extend, modify, or constrain the operational freedoms that the SPRIND-Freiheitsgesetz of December 2023 established. The statutory evaluation has been completed, the Fachaufsicht evaluation has been tendered, and the April 2025 coalition agreement commits the new government to strengthening SPRIND while placing all spending under Finanzierungsvorbehalt. A parliamentary inquiry filed in December 2025 has signaled that this evaluation will be substantive rather than ceremonial. Decisions will be made with or without an evidence base. The second challenge is European advisory credibility. The Netherlands, France, and others are designing breakthrough innovation agencies now, looking to SPRIND for design guidance, with France's Ministry of Economic Affairs already scoping an ARPA-style agency in coordination with SPRIND. Both audiences require the same input, namely a systematic, evidence-based account of which ARPA design features actually drive outcomes, which transfer to a European institutional context, and which require structural adaptation. That account does not currently exist.

The two existing approaches to answering these questions are inadequate. Anecdotal evidence drawn from US ARPA successes (the internet, GPS, mRNA vaccines) makes a compelling legitimation narrative, but individual success stories cannot establish which design features produced those outcomes, the public record systematically overweights successes, and SPRIND is not in fact replicating any single US ARPA. It operates at different TRLs, with entrepreneurial performers, using equity instruments no US ARPA has. Without decomposing the model into testable components, anecdotal evidence cannot answer the questions parliamentarians and sister agencies are asking. Surface benchmarking on visible features (PM model, challenge format, flat hierarchy) treats the US ARPA model as a monolithic template. It fails to test which features function because of US-specific context, including the approximately $886 billion defense budget, Other Transaction Authorities that bypass standard procurement, and dense venture capital, and which function independently of it. Without this distinction, European governments risk copying surface features while missing the mechanisms that matter, or dismissing features that would work with proper adaptation.

This PAE addresses three research questions. Q1: Where does the ARPA bundle work at SPRIND? Q1 tests implementation and serves the domestic legitimation challenge. Q2: What transfers from the US model and what required adaptation? Q2 separates universal features from justified adaptations and serves the EU advisory credibility challenge. Q3: Where is SPRIND falling short, and which gap most constrains breakthrough outcomes? Q3 identifies the binding gap and serves both audiences. Together they provide the evidence base for the recommendations. The client is SPRIND GmbH. The faculty advisor is Carolyn Fu of Harvard Business School.