If you run a research park, a tech transfer office, or a regional development program, you know the slide. Startups formed this year. Dollars raised. Maybe one company that made the news. It is the slide boards ask for, and it is easy to build.
Michael Fritsch, an economist at Friedrich Schiller University Jena, has written extensively on new business formation and regional development. In a recent paper in The Journal of Technology Transfer, he argues that the slide is measuring the wrong thing (Fritsch, 2026). Specifically: startup counts are an input. The outcome a region cares about is whether the region itself does better, and that takes years to show up.
This is a deep dive on that paper: what it argues, how it makes the case, where it is thin, how it lines up with other research on university commercialization, and what I think it means for people who build and run research parks.
The paper in one paragraph
"Struggling with entrepreneurial ecosystems" was published online on October 31, 2025, and appears in volume 51, issue 5 of the journal. It is open access. Fritsch takes on the entrepreneurial ecosystem idea, which he notes has gained enormous popularity in recent years, and says it has not yet delivered on three things: finding causal relationships, building theory, and giving policymakers usable guidance. He then proposes four fixes. Judge an ecosystem by regional performance, with long time lags. Count all new businesses, not only high-growth ones. Count established firms too. And treat the entrepreneurial ecosystem as one part of the regional innovation system rather than a separate concept.
How the argument is built
This is a conceptual paper, not an empirical study. Fritsch states plainly that no datasets were generated or analyzed for it. There is no sample, no regression, and no new survey. Instead, he works through the existing literature section by section: what the outcome variable should be, what counts as entrepreneurship, how ecosystems relate to innovation systems, what the key elements are, and what makes a set of elements behave like a system. It closes with a list of conceptual requirements and open research questions. The paper went through peer review, and the acknowledgments thank anonymous referees.
That matters for how to read it. The claims are arguments supported by prior evidence that Fritsch cites, not new findings. Where I quote numbers below, they are numbers Fritsch reports from other studies.
Finding 1: startup activity is an input, not the result
The core move is simple. If the reason anyone cares about entrepreneurship is regional growth, prosperity, and well-being, then measures of entrepreneurial activity are a means to that end. In his words, those metrics "represent 'only' input to achieve these development goals" (Fritsch, 2026, Section 2).
As a result, much of the ecosystem literature, which studies what drives new business formation, only indirectly answers the question that matters. Fritsch adds a second point that every board member should hear: the effect of a change in entrepreneurial activity can take a considerable period to appear in regional numbers, so any policy assessment has to allow for long time lags.
My read: a one-year scoreboard of startups formed is closer to a count of seeds planted than a measure of the harvest.
Finding 2: the high-growth startup is a rare species
Most ecosystem work narrows its attention to "productive" entrepreneurship, meaning highly innovative, high-growth startups. Fritsch gives three reasons that focus misleads.
- It is one model, not the model. The Silicon Valley pattern is one way entrepreneurship shapes a region. In many regions, growth comes from dynamic populations of small and medium-sized firms focused on customer-oriented innovation. He points to the German Mittelstand as an example.
- The numbers are tiny. Fritsch reports that one US study classified only 0.06% of all startups from 1988 to 2010 as showing significant growth, defined as going public or being acquired at a meaningful positive valuation within six years. He also notes that in countries such as Germany, the UK, and the US, less than one in a thousand startups is promising enough to attract venture capital.
- Winners cannot be picked early. Success is nearly impossible to predict at the start, and many high-growth firms do not expand until well after their founding. A pick-the-winners policy is not feasible.
Then the line that should land hardest in a rural state: Fritsch observes that some less populated regions rarely have a high-impact startup at all, which shows how limited a policy built around those firms is for them.
"High-performance start-ups do not emerge out of thin air, but are usually embedded in the regional economy." (Fritsch, 2026, Section 3)
Finding 3: established firms are part of the system
This is the section I found most useful. Fritsch argues that incumbent firms belong in any honest picture of regional entrepreneurship, for two reasons.
First, incumbents shape the pool of future founders. He notes that most founders of successful startups are between 40 and 50 years old and worked at an established firm before starting their own. Founders also tend to locate close to where they already live. The knowledge and success of incumbent firms shape the performance of the spinoffs their former employees start.
Second, much of the effect of new businesses on a region runs through competition with incumbents. Some incumbents lose demand. Others respond by getting more productive. Fritsch says these indirect effects are often larger than the growth of the newcomers themselves, and that even startups that fail soon after entry can push incumbents to improve products and processes.
Put plainly: the machine shop down the road and the regional manufacturer are not background. They are where the next founders learn their trade, and they are the firms whose response decides whether a wave of startups helps the region at all.
Finding 4: merge the ecosystem with the innovation system
Fritsch argues that entrepreneurial ecosystems and regional innovation systems share the same population, workforce, firms, schools, knowledge base, institutions, and policymakers. He finds it surprising that much of the ecosystem literature barely mentions innovation systems. His proposal is to treat the entrepreneurial ecosystem as a subsystem of the broader innovation system.
That raises a practical question he leaves open on purpose: should one agency handle both entrepreneurship and innovation policy, and if there are two, how should they coordinate?
What the paper does not settle
Fritsch is candid that the field has more questions than answers. A few he names directly:
- Which elements matter, causally. Widely cited lists of ecosystem elements, such as formal institutions, culture, networks, physical infrastructure, finance, leadership, talent, knowledge, demand, and producer services, are plausible. But he says empirical proof of a causal link from those elements to entrepreneurial activity and regional growth is largely missing, and simple correlations are not enough.
- What has to be local. Does venture capital or specialized knowledge need to sit inside the region, or can it come from outside? He notes the answer depends on a country's geography, and contrasts the large distances between US metro areas with Germany, where a major city is within an hour or two by car from almost anywhere.
- How to measure "systemness." Counting links between actors captures the static side. The ability of a regional system to reorganize after a shock is much harder to assess in advance.
- When public money is justified. An element being important and missing is necessary but not sufficient. The sufficient condition, in his framing, is that the private sector will not supply it within a reasonable time.
Limits worth naming
Here is where I would push back, or at least read carefully.
- No new evidence. The paper is an argument built on other studies. It tells you what to measure, not what the measurement will show.
- A European vantage point. Many of the examples come from Germany and the Mittelstand. Fritsch himself flags that US geography differs, and that matters for a state like South Dakota, where distances between cities are long.
- No ready metric. He says regional performance should be the target, with long lags, but he does not hand practitioners a specific indicator set. Every region will have to build its own.
- Breadth has a cost. My concern, not his: counting every new business and every incumbent is more accurate, but it is harder to manage to. A program can lose focus if everything counts.
Fritsch closes by saying the concept "is at a rather early stage of development" and "has remained largely descriptive" (Fritsch, 2026, Section 8). That is an honest place to land, and it is a reason to treat his four requirements as a checklist for better questions, not a finished model.
How it connects to tech transfer research
The same input-versus-outcome problem shows up inside the university, one level down.
Disclosure counts. A recent working paper by Pelaez et al. (2026), posted on SSRN and not yet peer reviewed, studies invention disclosures at a major US engineering university whose disclosure volume rose starting in 2022. Matching internal disclosure records to publications and using a measure of commercial potential, the authors find that inventions with higher commercial potential were much more likely to be disclosed, and that this sorting became substantially stronger after 2022, when volume rose. Their conclusion mirrors Fritsch's: disclosure volume alone is an incomplete indicator of tech transfer performance. They also note that the bundled nature of the changes at that university keeps them from isolating which change drove the result.
Office capacity. Swamidass and Vulasa (2009) analyzed AUTM licensing surveys from 1995 to 2004 and found licensing income was 1.7% of total research expenditure in 1995 and 2.9% in 2004. They then sent a survey to 99 randomly selected US research universities. Seventy-five percent of respondents reported staff shortages for legal and non-legal processing of inventions, and more than a third said they failed to process more than 26% of inventions in 2006 because of limited capacity. Their argument is that understaffed offices default to filing patents at the expense of marketing inventions. In Fritsch's terms, patent filings are a countable input; whether an invention reaches a firm that can use it is closer to the outcome.
Faculty and culture. Owen-Smith and Powell (2001) used fieldwork on two campuses to explain why disclosure rates differ so much. Faculty decisions to disclose depend on how they see the benefits of patent protection, the perceived cost of working with the tech transfer office, and whether the campus environment supports or opposes pursuing academic and commercial work together. That is the campus-level version of what Fritsch calls informal institutions, or regional culture, which he says changes slowly and deserves far more study.
Read together, the papers make one point at three scales. The campus, the office, and the region all reach for the number that is easy to count. The thing that matters sits further downstream and takes longer to see.
What this means for research parks and tech transfer offices
Thel Consulting works on tenant recruitment for the USD Discovery District, and we author the Mapping the Growth Potential market study. As a result, I spend a lot of time on the questions this paper raises. Here is my understanding of what it implies in practice.
- Report a second scoreboard. Keep the startup count, but pair it with regional measures tracked over years: employment, wages, and productivity in the sectors the park serves. Tell the board up front that the lag is real.
- Recruit established firms on purpose. If most successful founders worked at an established firm first, then a regional manufacturer or device company in the park is a founder pipeline, not just a rent check.
- Serve everyday entrepreneurs. The supplier, the contract machinist, and the services firm are part of how deep-tech companies get built in a region. Space, programs, and introductions should not be reserved for companies that look venture-backable on day one.
- Count licenses to local incumbents. For a tech transfer office, a license that helps an existing regional firm compete may matter as much as a spinout. Fritsch's point about incumbent response suggests it should be tracked and reported, not treated as second prize.
- Coordinate the two policy tracks. If entrepreneurship and innovation programs sit in different offices, the paper is a good reason to put them on one plan with shared measures.
- Be specific about what is missing. Before asking for public money for a lab building, a fund, or a program, show that the private market will not supply it in a reasonable time. That is Fritsch's sufficient condition, and it is also a stronger pitch.
My desired outcome for a park president or research VP reading this: the next annual report still has the startup slide, but it sits next to a slide about how the region is doing, and the second slide is the one the strategy answers to.
Related reading
- Discovering the Future of Research Parks: Flexible Master Planning + Strategic Marketing
- 2025 South Dakota Bioscience Market Study: Mapping the State's Growth Potential
- Founders want compute and the bench in one building
- 2026 Governor's Giant Vision Competition
References
Fritsch, M. (2026). Struggling with entrepreneurial ecosystems. The Journal of Technology Transfer, 51(5), 3083–3095. https://doi.org/10.1007/s10961-025-10280-0
Owen-Smith, J., & Powell, W. W. (2001). To patent or not: Faculty decisions and institutional success at technology transfer. The Journal of Technology Transfer, 26(1–2), 99–114. https://doi.org/10.1023/A:1007892413701
Pelaez, S., Yang, J. J., Ceccagnoli, M., & Walsh, J. (2026). More disclosure, better inventions? Commercial potential in university invention disclosures [Working paper]. SSRN. https://doi.org/10.2139/ssrn.6968389
Swamidass, P. M., & Vulasa, V. (2009). Why university inventions rarely produce income? Bottlenecks in university technology transfer. The Journal of Technology Transfer, 34(4), 343–363. https://doi.org/10.1007/s10961-008-9097-8
