Here’s a useful article that looks at the fate of university-licensed startup (ULS) life sciencecompanies over the last few years. There are more and more such companies (a greater than tenfold increase in their number since 1990), but a comprehensive look at success rates (and how such rates vary according to the universities involved) has been harder to come by. A lot of work went into this overview, and I’m glad the authors were able to assemble it. They spent a lot of time digging through tech-transfer records and contacting universities to make sure that they’d covered everything, checking venture capital announcements and state corporate registrations, and searching through press releases, LinkedIn pages, and more. The data that they have produced look like the best we have on the subject.
So here’s the overall picture. If you take the fifty top patent-producing schools over the last fifty years as your cohort, those are responsible for about 70% of all university patents with a long-tail distribution thereafter. Formation of new life science companies is, if anything, even more top-loaded. Only 47 of those 50 schools produced any such new firms at all during the 1980-2013 period. The top five universities produced one-third of the new companies, and the top ten produced half, while the bottom seven school produced 3% of them.
What fates did all these companies meet? 13% of them were eventually acquired by another company, and 10% of them had an IPO, so those are the successes. There’s an ambiguous category after that, of companies that have at least one entry in the Dun & Bradstreet Data Universal Numbering System, which is probably the most comprehensive registry of firms that appear to be (or have been) going concerns. 30% of the companies fit that, but with no apparent acquisition or IPO. Meanwhile, 18% of them show up in DUNS but appear to have only one or two employees, 22% show up but have either definitely failed or cannot be proven to still exist, and 7% never applied for a DUNS registration at all. Overall then, a generous reading would be roughly 23% success, 47% likely or definite failure, and 30% “work in progress”.
Interestingly, the number of ULS startups that seem to be active but show little evidence of actual progress has been increasing over the years. The authors have no problem naming what they’re seeing:
The walking dead represent symbolic action by universities, a transfer that made the TTO and others look good in the short term, the time horizon of annual budget fights and resource allocation decisions. Over the long term, however, no real business or economic development occurred. We saw the significant rise in the number of these firms as driven by university administrators attempting to game the system, to claim, but not create, economic development.
I find that very plausible indeed, unfortunately. As an illustration, the paper notes that in 2008 the two universities that (on paper, at least) formed the most startup companies were MIT and the University of Utah. Sadly, it turned out that the majority of the Utah companies listed the university’s tech transfer office address and phone number as their corporate headquarters and its director as a corporate officer, neither of which are signs of a successful spinout. And neither is this sort of behavior (which I would call “performative technology transfer”) peculiar to Utah.
Meanwhile, the venture capital firms seem to be doing a reasonably good job with their dollars: VC activity was very disproportionately centered around those companies that were eventually acquired or had an IPO. You can regard some of that as self-fulfilling prophecy, but at the same time, VC firms are also going to vet companies for their chances of achieving such exits before they invest at all, and they are definitely not going to be able to take a random sample of firms and achieve those same numbers. Not even in the current funding environment!
Finally, geography. The authors have location data covering 1990 to 2011, with 498 companies represented. 23% of those were founded in the well-established biotech clusters around Boston/Cambridge, the SF Bay area, or San Diego. Overall, 68% of those 498 companies stayed within 60 miles of the founding university, while the rest migrated. Firms founded in those clusters experienced success twice as often and failure half as often – there’s a reason that these things come to exist and to perpetuate themselves. The paper’s study of companies that have moved in and out of various locations suggest, as one would well believe, that it’s not so much that being in such a cluster makes a company successful as that more successful firms are more likely to form in such places in the first place. “Assortative matching” best fits their results – the idea that firms succeed because they have access to the resources that they need, and that they move in order to maximize these fits. Indeed, the lowest success rates were found among those firms that started off in a non-cluster location and never moved at all.
Overall, the authors suggest that the current system’s incentives, at least in many locations around the country, are misaligned. University administrators and state legislators are hoping for new economic activity and the benefits thereof, but they reward startup formation itself as if it were necessarily connected. And it isn’t. The “walking dead” examples mention show just where such a disconnect forms. Tech transfer offices should report better data, and should not be allowed (by universities, but also by outside observers) to simply report numbers of new firms “formed” or technologies “licensed out” without context. Longer time horizons for such things would be a good first step: has anything actually come of all these announcements?
The paper also points out that emphasizing economic development and achieving investment returns are two primary goals of university tech transfer offices, but that these may well be incompatible. Many IPOs, for example, provide employment and wages but also have negative returns for investors, whereas out-and-out acquisitions can show the opposite: little extra economic activity once the small company is swallowed up, but definite returns on the initial investment. Tech transfer offices should realize this, be more open about it, and make up their minds about which goals they find more important.
Finally, the authors have an interesting recommendation. They say that the two most likely ways for a region to fail in building a life-science ecosystem are either launching things into a vacuum and trying to compete with the big clusters (but just on a smaller scale). Instead, the paper argues, smaller areas would do well to pick some particular niche and try to become excellent in that area. Boston/Cambridge and the Bay area can fling startups all over the place – oncology, diagnostics, technology platforms, gene therapies, neuroscience. Try to replicate that from scratch and you’re going to be in trouble. But on the other hand:
Assortative matching implies that becoming a ‘micro cluster’ — a community with sufficient and well-developed resources that support a specialized technology or product area — should be sufficient to help local firms prosper and invite in-migration. Communities should not focus on being good in the ‘life sciences’; they should create and develop resources around some narrow slice of the sector and become best-in-class in that area. Universities should focus less on building strong colleges (for example, life science or medicine), and specialize more on outstanding departments (for example, genetics or ophthalmology).
We’ll see if anyone takes them up on the recommendation. In the meantime, I can strongly recommend this paper to anyone interested in the whole topic of university-based startups. There’s a lot of good information here that’s hard to come by anywhere else.