Ecosystem Governance: How National Innovation Agencies Are Built, Funded and Judged

Behind every startup ecosystem that governments admire sits an agency — or, more often, a small family of them — that made a set of unglamorous decisions about mandate, money, staff and accountability twenty years earlier. Israel’s innovation authority, Singapore’s enterprise agency, Britain’s Innovate UK, Finland’s Business Finland, Thailand’s science agency, India’s Startup India machinery and the Gulf’s new generation of enterprise, venture and technology bodies all descend from those decisions. This article compares their anatomy, reviews what the evidence says about the instruments they deploy, and sets out the governance principles that distinguish an agency that builds an ecosystem from one that announces it.

The idea in brief. National innovation agencies succeed when they have a clear mandate, operational independence with political accountability, staff drawn from industry and investment, a portfolio logic that tolerates individual failure, and an evaluation discipline that moves money towards what works. The instruments with the strongest evidence are direct R&D grants to young firms, tax incentives for research, and fund-of-funds structures that crowd in private managers rather than replacing them; pure government venture investing performs worst. India’s and the Gulf’s agencies are young by these standards and are converging — India from scale towards depth, the Gulf from capital towards pipeline — on the same governance model. The test of any agency is not the size of its announcements but the cost per outcome and the private capital it crowds in.

Why agencies at all: the argument, briefly

The case for public innovation agencies rests on a market failure that economists have accepted for seventy years: firms under-invest in research and early-stage risk because they cannot capture all the returns, and young firms in particular cannot finance experiments whose value is mostly information. The argument’s modern statements sit on two shelves. Mariana Mazzucato’s The Entrepreneurial State made the maximal case — that public agencies have been the patient, risk-taking investors behind many of the technologies private firms later commercialised. Josh Lerner’s Boulevard of Broken Dreams (Princeton University Press, 2009) catalogued the far larger number of public venture and entrepreneurship programs that failed, and distilled why: political capture, poor design, no evaluation, and a preference for announcements over follow-through.

Dani Rodrik’s contribution was to change the question. Industrial policy, he argued, should not be debated as whether but as how: governments will intervene regardless, so the useful work is designing institutions that discover what works and stop what does not. Réka Juhász, Nathan Lane and Rodrik’s 2023 survey of the new empirical literature, published as an NBER working paper, finds that well-designed interventions can produce measurable gains and that design and governance explain most of the variation in results. Nicholas Bloom, John Van Reenen and Heidi Williams’s 2019 toolkit review in the Journal of Economic Perspectives ranked the instruments by evidence: research tax credits and direct grants have solid support, skilled immigration is strongly supported, and entrepreneurship-specific interventions have thinner and more mixed evidence — a warning that agencies in this field are operating with less certainty than their brochures suggest.

Anatomy of the exemplars

Two historical cases repay closer reading. Israel’s Yozma program of 1993 — a government fund of roughly US$100 million that co-invested in new private venture funds on terms that let the private partners buy out the state’s share cheaply if the funds succeeded — is the most-cited example of a public instrument that created a private industry and then withdrew. Gil Avnimelech and Morris Teubal’s 2006 analysis in Research Policy placed Yozma within a longer sequence of policies that had built the demand for venture capital before supplying it: R&D grants first, then incubators, then the fund-of-funds. The order mattered. Dan Breznitz and Darius Ornston’s 2013 comparison of Finland and Israel in Comparative Political Studies added an institutional insight: radical policy innovation came from small, peripheral, low-profile agencies with room to experiment, not from the powerful ministries at the centre of government. Agencies that are too important to fail are also too important to try anything.

The instrument menu, and what the evidence says about each

The fund-of-funds deserves a closer look because it is the instrument every corridor government now holds. India’s Fund of Funds for Startups, launched in 2016 with a corpus of ₹10,000 crore managed by SIDBI, invests in SEBI-registered venture funds rather than in companies; a further fund of the same size was announced in the 2025 Union Budget. Saudi Arabia’s Jada, established in 2018 under the Public Investment Fund with capital in the region of four billion riyals, invests in venture and private-equity funds; the kingdom’s Saudi Venture Capital Company invests in funds and co-invests directly. Bahrain’s Al Waha fund-of-funds, launched in 2018 through the Bahrain Development Bank at around US$100 million, was small but pointed. The chart compares announced sizes; the more important comparison — private capital crowded in per public dollar, and the quality of the managers backed — is one each agency should publish.

Yozma’s lesson applies to all of them: the fund-of-funds works when it builds a private industry that outgrows it, and it should be designed with its own obsolescence in mind — terms that reward private partners for success, a sunset, and a measured exit.

Governance principles that separate builders from announcers

1. A clear, narrow mandate. The exemplars do one or two things. Agencies given investment promotion, SME lending, startup programs, research funding and skills in one body do all of them at the pace of the slowest.
2. Independence with accountability. Operational decisions — which firms, which funds, which programs — insulated from political direction; strategic priorities and outcome reporting fully accountable to the ministry and legislature. Breznitz and Ornston’s peripheral agencies had this room; the ministries around them did not.
3. A board and staff from the market. Investors, founders and industry executives on the board; deal-makers, program operators and analysts on the staff — paid enough to hire them, and rotated so that the agency does not become a career refuge.
4. Portfolio logic. Most funded ventures will fail; the agency’s return comes from the few that do not. A governance culture that punishes individual failures produces an agency that funds nothing risky and therefore nothing important.
5. Additionality as the test. Every instrument should be able to answer: what happened that would not have happened otherwise? Where oversubscription exists, the agency should use it to find out (article 32).
6. Sunset and evaluation. Programs with end dates and external evaluations; instruments that are designed to be withdrawn when the market they were meant to create exists.
7. Coordination as a product. Where several agencies share a jurisdiction — as in Bahrain and Saudi Arabia, or between India’s centre and states — a single front door for the founder, investor or company is worth more than any single instrument.

Judging an agency

Agencies are judged, in practice, by the numbers they choose to publish, and the choice reveals the governance. India’s Startup India registry offers a clear example of a number that is both genuinely impressive and easily misread.

The registry’s growth — from a few thousand recognised startups in its first full year to well over one hundred and fifty thousand by early 2025, according to the Department for Promotion of Industry and Internal Trade — reflects real formalisation: firms that registered because recognition unlocks tax and compliance benefits, and because being counted is now normal. It is a measure of reach, not of outcome. An agency judged on this line alone would be judged on its own marketing. The numbers that judge an agency are the ones set out in article 32, adapted to the instrument:

• Cost per outcome — per venture operating at 24 months, per licence executed, per learner placed, per fund raised — by cohort and instrument.
• Additionality — the difference between supported and comparable unsupported firms, from lottery, staggered rollout or matched designs.
• Private capital crowded in per public unit, for fund-of-funds and co-investment instruments, with the managers named.
• Time to service — days from application to decision, and from decision to money, the numbers founders actually compare across countries.
• Portfolio results — survival, growth and exits across everything funded, reported with the failures included.
• Instrument retirement — which programs the agency has closed, and why. An agency that has never stopped anything has never evaluated anything.

The Indian and Gulf trajectories

India built breadth first. The registry, the incubation network, the fund-of-funds, the research and skills missions, and a dense set of ministry-level programs have given it the largest startup-policy apparatus in the developing world, and its states now compete with agencies of their own. The current frontier is depth: outcome measurement across the apparatus, evaluation of which programs move ventures rather than register them, coordination between the centre and states so that the founder sees one system, and a fund-of-funds discipline that publishes crowd-in and manager quality. The scale is India’s asset; the governance of that scale is its task.

The Gulf built instruments first. In half a decade Saudi Arabia assembled an SME agency, a venture-fund investor, a fund-of-funds, an investment ministry and a technology-development program; Bahrain has run its three-agency model for two decades; the UAE’s emirate-level platforms and free zones give it several ecosystems within one country. Capital is not the constraint. Pipeline and coordination are: enough founders, researchers and firms to absorb the instruments productively, a single front door across agencies, and — as the instruments mature — the evaluation discipline that decides which to grow and which to retire. The corridor’s complementarity (article 36) is visible here too: India’s pipeline and program depth, the Gulf’s capital and coordination capacity, each what the other lacks.

The state and emirate layer

National agencies are no longer the whole story on either side of the corridor. India’s states run their own startup missions, incubation networks and seed funds — Karnataka, Telangana, Kerala, Gujarat, Tamil Nadu and Maharashtra among the most active — and compete for founders and capability centres with policies of their own; the UAE’s emirates and Saudi Arabia’s regional authorities are building comparable layers. The sub-national layer is healthy when it competes on execution — time to service, program quality, landing support — and wasteful when it duplicates national instruments: a state fund-of-funds that backs the same managers as the national one, a registry that recognises the same startups twice, an incubation grant that funds a building the national mission already funded. The governance answer is a division of labour written down: national bodies own the instruments that need scale and consistency — registries, tax treatment, fund-of-funds, research funding — and sub-national bodies own the instruments that need proximity — programs, landing services, aftercare, campus and research links. The founder, in either case, should see one front door.

A composite case: the agency with three mandates and no portfolio

A composite from several engagements; details altered.

A national agency in the region had been assigned investment promotion, startup programs and SME finance in successive decrees. Each mandate had a division; each division had a dashboard; none had a portfolio view. The startup division ran programs whose graduates the finance division did not lend to; the promotion division attracted companies that neither division served after landing. A review commissioned by the cabinet asked the agency for a single statement of outcomes across all three mandates and found that the question could not be answered because the divisions defined success differently and shared no data.

The reform narrowed and connected. Investment promotion moved to a dedicated body with its own aftercare function. The remaining agency adopted one mandate — enterprise development — with a board drawn from investors and founders, a portfolio view across programs and finance, twelve outcome KPIs with definitions and horizons, and an evaluation unit reporting to the board. Its first act was to retire two programs whose outcomes could not be evidenced and to redirect their budgets into the fund-of-funds and a cohort accelerator whose numbers could. Its second was to publish, for the first time, a portfolio report that included failures. The cabinet’s review question was answered eighteen months later, in one document, with a cost per outcome on the cover.

What could go wrong

• Mandate accretion. Every new priority becomes a division. Antidote: narrow mandates; new bodies for new missions.
• Political selection. Ventures and funds chosen for who they know. Antidote: independent investment committees; published criteria; private co-investment as a condition.
• Government as sole investor. The weakest instrument in the evidence. Antidote: fund-of-funds and co-investment that require private capital alongside.
• No exit design. Instruments outlive the market failure they addressed. Antidote: sunsets; Yozma-style buy-out terms.
• Registry as result. Recognition counts reported as ecosystem outcomes. Antidote: outcome KPIs by instrument.
• Staff without market experience. Antidote: pay and rotate for investors, founders and operators.
• Evaluation as threat. Programs never assessed because assessment might close them. Antidote: an evaluation unit with a mandate to retire programs, reporting to the board.

Questions ministers actually ask

“Should we create a new agency or reform the old one?” Reform if the mandate can be narrowed and the board rebuilt; create if the new mission would otherwise become another division. Small, peripheral bodies have the best record of innovating.

“Which instrument first?” In the order that builds demand before supply: grants and programs that create investable ventures, then a fund-of-funds that crowds private managers into financing them, then the instruments that need a functioning market to work.

“Should the state invest directly in startups?” The evidence says no. Invest in managers, co-invest alongside private capital, fund the programs that produce the pipeline — and let private investors pick the companies.

“How do we compare with Israel or Singapore?” On governance, not on size: mandate clarity, independence, market staff, portfolio logic, evaluation and retirement. Their instruments can be copied; their governance is what made the instruments work.

“What should our first annual report contain?” Cost per outcome by instrument, additionality where measured, private capital crowded in, time to service, portfolio results including failures, and the list of what was retired.

Methodology & data notes

Agency founding dates, mandates and instrument sizes are as published by the agencies and their governments and are approximate where currencies are converted; fund sizes are announced corpuses, not deployed capital, and readers should consult the agencies for current figures. Findings from the academic literature are summarised approximately. Startup India registry counts are cumulative recognitions as published by DPIIT and are a measure of reach rather than outcome, as the text explains. The composite case combines several engagements with details altered. Companion articles cover ecosystem measurement (article 32), founder programs (article 31), incubator economics (article 37), investment promotion (article 35) and the India–Gulf corridor (article 36).

References & further reading

• Lerner, J. (2009) — Boulevard of Broken Dreams: Why Public Efforts to Boost Entrepreneurship and Venture Capital Have Failed — and What to Do About It, Princeton University Press
• Mazzucato, M. (2013) — The Entrepreneurial State; UCL Institute for Innovation and Public Purpose
• Juhász, R., Lane, N. & Rodrik, D. (2023) — The New Economics of Industrial Policy, NBER Working Paper 31538
• Bloom, N., Van Reenen, J. & Williams, H. (2019) — A Toolkit of Policies to Promote Innovation, Journal of Economic Perspectives
• Avnimelech, G. & Teubal, M. (2006) — Creating venture capital industries that co-evolve with high tech: Insights from an extended industry life cycle perspective of the Israeli experience, Research Policy
• Breznitz, D. & Ornston, D. (2013) — The Revolutionary Power of Peripheral Agencies: Explaining Radical Policy Innovation in Finland and Israel, Comparative Political Studies
• Howell, S. (2017) — Financing Innovation: Evidence from R&D Grants, American Economic Review
• Lerner, J. (1999) — The Government as Venture Capitalist: The Long-Run Impact of the SBIR Program, Journal of Business
• Brander, J., Du, Q. & Hellmann, T. (2015) — The Effects of Government-Sponsored Venture Capital: International Evidence, Review of Finance
• Israel Innovation Authority, Enterprise Singapore, Innovate UK, Business Finland and NSTDA — agency sites
• Startup India, Atal Innovation Mission and SIDBI — India’s startup-policy apparatus
• Bahrain EDB, Tamkeen, Saudi Venture Capital Company and Jada — Gulf agencies and fund-of-funds (Monsha’at, the Saudi SME authority, publishes on its own portal)

HexGn advises ministries and agencies on ecosystem governance — mandate design, instrument sequencing, program architecture and the measurement that lets an agency publish a cost per outcome — and delivers the founder, campus and research programs that give the instruments a pipeline.