Every innovation agency eventually builds a building. The logic is irresistible: a capital budget is easier to secure than an operating one, a building can be opened by a minister, and occupancy is a metric that rises. Two years later the building is full, the tenants are pleasant, and the agency cannot name a company that is bigger because it was there. This is the real-estate trap, and most of the world’s publicly funded incubators are in it. This article explains how the trap works, what the research says about the designs that escape it, and what a government should specify when it commissions an incubator or accelerator that is meant to change outcomes rather than fill floors.
The idea in brief. Incubators and accelerators are program businesses that are frequently financed as property businesses, and the financing shapes the behaviour: rent-dependent operators select for tenants who pay, not ventures that grow. The research is clear that what changes outcomes is the program — selection, intensity, structured mentoring, demand-side connections and a time limit — and that the space is incidental. Because a small share of young firms create most new jobs, selection and follow-through matter more than capacity. Agencies should fund operators on outcomes, cap the share of revenue that can come from rent, and design the accelerator as a cohort program with a building attached rather than a building with a program attached.
The real-estate trap, explained
The trap has three mechanisms, and understanding them is the first step to designing around them.
Capital is easier than operating money. Public budgeting treats a building as an asset and a program director’s salary as a cost. A ministry can justify a one-time capital grant for an innovation centre far more easily than a decade of operating funding for coaching, mentoring and follow-up. The building gets built; the program that would make it useful is funded for a year, then two, then reviewed.
Rent is the only revenue that arrives monthly. Once operating money is scarce, the operator discovers that desks, offices and event space generate predictable income and that programs do not. Within a few budget cycles the operator is a landlord who runs events, and its selection criterion has quietly become the ability to pay rent — which selects for consultancies, service firms and comfortable small businesses rather than the risky young companies the centre was built for.
Occupancy is the metric that rises. Because the building is the visible investment, occupancy becomes the reported KPI, and occupancy rewards exactly the wrong tenants: stable ones who stay. A well-functioning accelerator has high churn by design — ventures arrive, are transformed or stopped, and leave. A landlord’s dashboard reads that churn as failure.
The trap is not a moral failing of operators. It is the predictable response to how the centre was financed. Escaping it requires changing the financing, not exhorting the operator.
What the research says about sponsorship and accelerators
The scholarly literature on incubation and acceleration has matured over the past fifteen years, and its findings point in one direction.
Alejandro Amezcua and colleagues’ 2013 study in the Academy of Management Journal, examining thousands of incubated firms, found that the benefit of organisational sponsorship — incubation — depended heavily on the founding environment: sponsorship helped most where resources were scarce and competition intense, and could even be neutral or negative where the environment was already rich. Incubation, in other words, is a substitute for a missing ecosystem, not a supplement to a present one — a finding with direct implications for where governments should locate centres.
Susan Cohen, Daniel Fehder, Yael Hochberg and Fiona Murray’s 2019 paper in Research Policy mapped the design of accelerators — the fixed-term, cohort-based, mentorship-driven programs that end in a demo day — and showed how much variation hides inside the label: in selection, in the intensity and structure of mentoring, in the role of the sponsoring organisation, and in whether the program takes equity. Their typology is the most useful starting point for any agency specifying a program, because it turns “accelerator” from a brand into a set of design choices.
The outcome evidence, discussed in article 31, adds the mechanism: Benjamin Hallen, Susan Cohen and Christopher Bingham’s 2020 study in Organization Science found that effective accelerators compress learning through intensive, structured consultation; Sandy Yu’s 2020 analysis in Management Science found that accelerated ventures resolve uncertainty faster — including by closing sooner when the idea does not work; and the Start-Up Chile evaluation found that space and cash without schooling did nothing measurable. Across these studies the building never appears as a variable that matters. The program does.
The vital few: why selection matters more than capacity
A second body of research explains why incubators should be selective rather than capacious. John Haltiwanger, Ron Jarmin and Javier Miranda’s 2013 study in the Review of Economics and Statistics established that it is young firms, not small firms, that drive net job creation in the United States — and that most young firms either fail or stay small, while a minority grow rapidly and account for the bulk of the jobs. Ryan Decker and colleagues’ 2014 review in the Journal of Economic Perspectives extended the point: entrepreneurship’s economic contribution is concentrated in a small share of high-growth ventures. Nesta’s much-cited British analysis, The Vital 6%, found that a small minority of firms generated more than half of net new jobs over the period studied.
For incubator design the implication is stark. If outcomes are concentrated in a few ventures, an incubator’s value lies in finding and intensively supporting those few — which argues for selective intake, deep support per venture, and follow-through after the program — not in hosting as many companies as the floor plate allows. A centre that supports two hundred tenants lightly is unlikely to contain the vital few; a program that supports twenty ventures intensively, selected with care and connected to demand and capital, might. The building’s capacity is the wrong unit of ambition.
Five operating models, and what each is for
| Model | Primary revenue | Selects for | Typical outcome | Best use |
|---|---|---|---|---|
| Real-estate incubator | Rent and services | Ability to pay | High occupancy; low venture growth | Managed workspace — call it that |
| University incubator | University budget; grants | Affiliation; research links | Continuity for student and research ventures; weak market pull | The continuation pathway for campus and research programs (articles 33 and 34) |
| Corporate accelerator | Corporate innovation budget | Fit with the sponsor’s needs | Pilots and procurement; risk of capture | Demand-side connection for a sector program |
| Government-funded accelerator | Public grant; sometimes equity | Policy priorities; readiness | Depends entirely on design and operator | National and sector founder pipelines, when run as a cohort program |
| Network or virtual program | Fees; sponsorship; equity | Founder quality; remote reach | Reach without density; variable depth | Tier-two cities and cross-border cohorts |
The models are not exclusive, and the strongest ecosystems contain all five. The error is to fund one model while expecting the outcomes of another — most commonly, to fund a real-estate incubator and expect the outcomes of a government accelerator.
The economics: cost per outcome, not cost per desk
The unit that should govern the design is cost per venture operating at 24 months — the same measure proposed for founder programs in article 31 and for ecosystem measurement in article 32. The chart below is an illustrative model, indexed to the space-led incubator, of how that cost typically compares across three operating models when program costs, occupancy, selection and follow-through are modelled honestly.
The model’s logic is simple. The space-led incubator has high fixed costs, broad and shallow intake and low conversion, so its cost per operating venture is high even though its cost per desk is low. The grant-led model spends less on fixed costs but, without intensive support, converts only modestly better. The program-led accelerator has higher costs per venture during the program but selects tightly and converts far better, so its cost per outcome is lowest. The numbers in any real case will differ; the ordering rarely does.
Three financing rules follow for public funders. Fund the program, not only the building: a multi-year operating grant tied to outcomes is worth more than a capital grant of ten times the size. Cap rent dependence: if more than a modest share of an operator’s revenue comes from tenants, its incentives have already changed; a cap written into the funding agreement keeps them honest. Pay for outcomes at the margin: a base grant for running the program plus a performance element for ventures operating, revenue and capital raised at 24
