You are the provost. Year five of a flagship research initiative, and you are standing in a parking lot at six in the evening, watching it empty. The patents are filing. The papers are publishing. The rankings are climbing. Not one company of consequence has spun out of any of it. The labs are full. The lot is not.
That pattern repeats across dozens of university towns worldwide, and it has puzzled economists, policy advisors, and jealous administrators for decades. Why does one mid-sized college city produce a cluster of venture-backed companies while a geographically comparable one, with a similarly ranked engineering school, produces mainly adjunct professors and good coffee shops?
Research output and startup formation are almost entirely different activities. A university can be excellent at one and structurally hostile to the other. That is not a diplomatic observation; it is a diagnosis.
The mythology of the research spillover
The standard story, still taught in regional economics courses, goes roughly like this: universities produce knowledge, knowledge spills over into the local economy, entrepreneurs commercialize that knowledge, a cluster forms. Stanford begat Silicon Valley. MIT begat Route 128. Build a great university and wait.
The story is mostly backward.
Both Stanford and MIT had something far more specific than research excellence: faculty who held equity in companies, administrators who actively encouraged professors to consult for industry, and a local culture that did not treat leaving academia for a startup as professional suicide. Frederick Terman at Stanford was not simply a dean. He was a recruiter, a connector, and an advocate who spent decades making the case that applied work was honorable. That normative shift preceded the economic one by a generation.
Compare that with a school where the tenure committee treats industry collaboration as evidence of insufficient seriousness, where the technology transfer office requires a two-year exclusivity negotiation before any license moves, and where the most celebrated faculty outcome is a named professorship. Same research output. Radically different commercial trajectory.
The three things that actually separate the towns
Set aside research rankings for a moment. Three structural factors do most of the explanatory work.
Equity culture among faculty. When professors can hold meaningful ownership stakes in spinouts without surrendering their university positions, the incentive calculus changes completely. A materials scientist who knows she can take a two-year leave, co-found a company, retain her lab, and return if the venture fails will behave very differently from one who must choose permanently between the university and the market. MIT's policies on faculty entrepreneurship have been explicitly permissive for longer than most competing institutions have even had formal technology transfer offices. That permissiveness compounds: it attracts entrepreneurially-minded faculty in the first place, which changes the department culture, which shapes what graduate students believe is possible. The flywheel effect is real, and it starts with a policy document almost nobody reads.
The density of patient, local capital. This one is underappreciated, and the numbers bear it out. Venture capital is not fungible across geography, at least not at the seed stage. An early-stage investor who can drive twenty minutes to a founder's lab, who knows the department head socially, and who has already backed two other companies from the same research group is operating with information a remote investor simply cannot replicate. The Research Triangle in North Carolina spent roughly two decades building research infrastructure before local capital density reached the threshold where it could self-reinforce. Towns that try to skip that accumulation phase by recruiting a single venture fund rarely see it take hold; a lone fund without a surrounding ecosystem is like one coral polyp in open water.
A permission structure for failure. This is the least quantifiable factor and probably the most important. In some university towns, a failed startup is a credential. In others, it is a mark. The difference shows up in social behavior: whether founders talk openly about what went wrong, whether local press covers failure with the same seriousness as success, whether the university itself celebrates the attempt rather than the outcome. Austin developed this culture over time, partly through deliberate institution-building (the IC2 Institute at UT Austin explicitly set out to normalize commercialization across the 1970s and 1980s) and partly through the self-selection of people who moved there because the culture felt permissive.
Two professors, one conference, different endings
Consider a scenario that plays out in variants constantly. Two biomedical engineers, call them Priya and Daniel, attend the same conference and each present research with obvious commercial potential: a low-cost diagnostic tool that could work in low-resource clinical settings.
Priya is at a university where her department chair has already co-founded two companies, where the tech transfer office runs a forty-five-day fast-track licensing process for faculty spinouts, and where a local seed fund has a standing relationship with her school. Within eighteen months she has a co-founder, a small check, and a prototype in a clinic.
Daniel is at a school with a comparable research budget and a higher ranking. His tech transfer office is thorough, cautious, and slow. His department regards commercialization as a distraction. The one local investor he meets wants a revenue-generating product before writing a check. Three years later, Daniel's diagnostic tool is a well-cited paper. Priya's is a company.
The research quality was equivalent. The institutional infrastructure was not. That gap is measurable, and it widens every quarter Priya's company operates while Daniel's sits on a shelf.
What people get wrong about replication
The standard mistake that policy-makers make, and they make it reliably, is to treat startup ecosystem formation as an amenities problem. Build a nicer coworking space. Recruit a famous accelerator. Announce an innovation district. Rename a parking lot. None of those things are worthless, exactly, but they address the symptom rather than the mechanism.
What actually needs to change is the incentive structure for the people who hold the knowledge: the graduate students who might become founders, the postdocs who might leave for a startup, the junior faculty watching their senior colleagues to understand what is rewarded.
Ask yourself: if you were a brilliant thirty-year-old with a promising idea and a choice of institutions, would you choose the one with the nicer coworking space, or the one where the last three people who tried something commercial still have their jobs and their reputations?
There is also a temporal illusion at work. MIT and Stanford look like overnight successes from a distance of seventy years. Up close, each took roughly two decades of deliberate institutional norm-setting before the flywheel turned. Administrators who expect five-year results from a ten-year project will always be disappointed, and will usually blame the wrong variables.
The towns that have pulled it off share one less-discussed trait: they stopped trying to copy Silicon Valley and started asking what their specific research strengths, their specific labor pool, and their specific industrial neighbors actually demanded. Pittsburgh built around robotics and medical devices because Carnegie Mellon's strengths and the legacy of the steel industry's collapse created specific, local demand for exactly those capabilities. It is not Silicon Valley. It was never going to be. It is something more durable: a cluster built on what was actually there, and that distinction is what makes it defensible when the next wave of capital moves on.
The provost staring at that empty lot is probably not running a bad university. The harder question is whether the institution has decided, at the level of policy and culture and tenure criteria, that it wants to be something else too. Until that decision is made explicitly, the parking lot will keep emptying at six.