This article originally appeared in American Affairs Volume VI, Number 4 (Winter 2022): 3–13.
How to Make an Entrepreneurial State: Why Innovation Needs Bureaucracy
by Rainer Kattel, Wolfgang Drechsler, and Erkki Karo
Yale University Press, 2022, 288 pages
Gone are the debates about limited versus big government. The motivating question across most of the political spectrum today is how to build an effective government. Or perhaps less ambitiously, how can we build a government that doesn’t immediately fail when it tries something new?
Liberals are frustrated that the United States can’t build public transportation or deploy clean energy fast enough. Conservatives are increasingly interested in industrial policy to counter a rising China. Congress just passed massive science funding, infrastructure, and climate packages and is banking on the administrative state to effectively implement them. A high-inflation macroenvironment has heightened the awareness of supply-side bottlenecks across the economy and turned dredging and port automation into headline grabbers. Public intellectuals like Ezra Klein and Derek Thompson are writing about the importance of a liberalism that builds and the need for an “abundance agenda.” Nearly everyone is disappointed with how our public health agencies handled the Covid-19 pandemic (even as they disagree about what went wrong).
These conversations are by no means limited to the United States, and the appropriately titled How to Make an Entrepreneurial State, by a trio of European academics—Rainer Kattel, Wolfgang Drechsler, and Erkki Karo—is the latest attempt to provide a handbook with some answers. Building on the work of Mariana Mazzucato’s The Entrepreneurial State from 2013, the book is a fiery defense of the essential role the public sector will need to play in supporting innovation to solve the great challenges of our age, and seeks to inform practical policy.
The book makes the important observation that state support for innovation is well and good, but without a clear plan for improving state capacity, such efforts often prove to be wasteful and ineffective. As it notes, “In innovation policy and beyond, we have decades of lessons and empirical proof that just focusing on and overemphasizing public sector agility, entrepreneurship and innovation does not work.”
The book’s lofty ambitions make its eventual shortcomings more disappointing, however. How to Make an Entrepreneurial State gets mired in high-level abstractions and historical case studies, and ultimately neglects practical questions. A better title for the book might be “Why to Make an Entrepreneurial State”—the “How” is scarce to be found.
The book is at its strongest when describing the evolution and transformation of various “innovation bureaucracies” across time and around the world. While digging into historical examples like the original Advanced Research Projects Agency, the Japanese Ministry of International Trade and Industry, and the Swedish innovation agency Vinnova, the authors surface certain themes that repeat across case studies: the need for flexible hiring rules, the importance of attracting a nation’s best and brightest into public service, and the significance of overarching “missions” to focus the public sector.
A unifying theme discussed throughout the book is the idea of “agile stability” for state bureaucracies: the tension between ensuring stability in the core functions that they provide to the public while also maintaining the flexibility to evolve and add new capabilities over time. Nevertheless, the book is frustrating in its lack of practical detail as to how agencies can embed agility into their functions or how to structurally enable bureaucratic actors to take risks. But the fundamental question is an essential one for our current moment, and the book can help point us in the right direction.
So, how does one actually make an entrepreneurial state? There is no single correct model, and in fact, an essential element of entrepreneurship is the ability to correct course and revise plans in real time to accomplish overarching goals. With that in mind, some recent attempts at rebuilding state capacity and fostering agile stability within the U.S. federal government can shed some light on the “how” of making an entrepreneurial state.
The authors often contrast two views of innovation, one stemming from Max Weber, a German sociologist and theorist of administrative bureaucracy, and the other from Joseph Schumpeter, the Austrian-born economist who coined the term “creative destruction.”
The authors make a distinction between Weber Type I and Type II organizations. A Type I organization is characterized by stability and process orientation in bureaucratic administration, whereas a Type II organization is typically smaller, more nimble, and willing to experiment with new models of service delivery or organization. The authors discuss organizations switching between the two models either at an individual level or at an ecosystem level, as nimble new agencies are started to meet new challenges (Type II) but eventually solidify into a “normal” Type I structure. The authors then theorize a new, blended model—classified as a Weber Type III organization—which is able to host both Type I and Type II capabilities under a single umbrella. This typifies the agile stability they consistently praise. (This kind of innovation jargon is unfortunately littered throughout the book, and I will hereafter attempt to avoid it.)
Schumpeter, meanwhile, is used somewhat as a foil, and his views on innovation through competition are dismissed as out of reach for public agencies, which need to maintain a higher degree of stability than the private sector. The authors write: “As Schumpeter argues, fear of death by competition is what drives businesses to innovate and explore new pastures. Companies sense opportunities and threats, seize them—or not—and reconfigure what they do. The public sector is not primarily defined by competition, even if in some cases we pretend otherwise and create rankings of schools, doctors, and so on, or compete with other countries, war-like or in peace.” But this may be giving short shrift to the suitability of creative destruction as a model.
At first glance, it is true that the concept of creative destruction has some obvious barriers to application in the public sector. For a variety of practical and political reasons, it would be difficult to set up a direct competitor to the Federal Aviation Administration, for instance, and then ask airlines to decide which agency they would rather be governed by. But that doesn’t rule out the idea of trying to foment the gale of creative destruction within public agencies, as opposed to across them.
To illustrate how this could work, we might look at (what I hesitantly call) “The Netflix Model of Institutional Competition.” For the first decade of its existence, Netflix was essentially a DVD-by-mail company. Not content with this business model, and intrigued by the early success of YouTube, leadership at the company started investigating the possibility of streaming video directly over the internet, and launched a streaming product in 2007. To do this, they set up a new division that was somewhat isolated from the traditional processes and metrics used to govern the rest of the firm. According to one history of the company, the DVD executives were even kicked out of certain management meetings to make sure the streaming division had room to grow.
This separate streaming division eventually grew to become the modern Netflix. Netflix competed with itself internally across product verticals rather than let an outside competitor beat it to the new market. But for this new team to have success, they needed to be able to pull from a different talent pool, be judged by a different set of metrics, and productively interface with the rest of the company without kickstarting a turf war. In other words, they needed a high degree of institutional autonomy.
A similar scenario might be playing out right now with the launch of two brand new science funding offices: the Advanced Research Projects Agency for Health (ARPA-H) at the National Institutes of Health (NIH) and the Technology, Innovation, and Partnerships (TIP) directorate at the National Science Foundation (NSF). ARPA-H, which is modeled after DARPA—the innovative defense research agency which helped push forward the internet, GPS, and mRNA vaccines—is tasked with seeking out the set of high-risk, high-reward medical breakthroughs that the NIH is often criticized for missing. Likewise, the new TIP directorate’s mission is to develop a new portfolio of applied technology investments, regional innovation efforts, and novel funding mechanisms.
Both ARPA-H and the TIP directorate have been granted a degree of institutional autonomy by Congress and have a clear mandate to push the envelope on funding research, identifying talent, and working with the private sector in ways that the parent agencies have been unable or unwilling to pursue. If ARPA-H and TIP succeed in pushing forward the technological frontier, and in piloting new ways to identify promising scientific investments, their parent agencies will inevitably feel pressure to begin scoping similar mechanisms or to step up their performance in other ways. Oftentimes, having a shining example of what good agency behavior makes possible is more productive for spurring reform than ten examples of agency failure. For instance, if ARPA-H is now able to successfully fund a younger cohort of scientists with positive impact, it might highlight the fact that NIH could be doing more to enable talented early-career researchers.
While it’s unlikely that ARPA-H and TIP will fully replace their host organizations in the same way that streaming replaced DVDs at Netflix, providing a bit of internal competitive pressure to our premier science funding agencies would be a welcome step toward a more entrepreneurial state. And this model of creating new agency divisions with a high degree of autonomy, which can eventually augment the agencies hosting them, could be more generally applicable.
If we want federal agencies to be able to embody agile stability or, more simply, to be able to learn how to improve their operating procedures over time, then we need to think carefully about the bureaucratic incentives to experiment. And by “experiment,” I don’t simply mean trying out a new system or process in a one-off fashion. I mean the scientific sense of experiment—systematically testing a hypothesis using careful design, data collection, and evaluation with an eye toward the likely (or, in the case of a randomized trial, the actual) counterfactual impact.
It may seem obvious in the abstract, but it is often difficult for federal agencies to prioritize a softer target like “run more experiments” unless there are specific individuals who see experimentation as a discrete goal tied to their career advancement. When an agency’s performance is judged primarily on short-term outcomes—how many vaccines were shipped, were the grants paid out, and so on—it’s quite easy for longer-term process improvements to fall by the wayside.
One way to incentivize such improvements is to build new offices or directorates within an organization (readers may be sensing a theme here) that have an explicit mandate, dedicated funding, and enough autonomy to run experiments on different aspects of the agency’s core mission. An actually existing example of this kind of experimentation center within the federal government is the Center for Medicare and Medicaid Innovation (CMMI), which is a division of the Centers for Medicare and Medicaid Services and works in precisely this way to pilot new health care delivery and payment models. For example, CMMI is the division tasked with actually testing a new payment model for hip replacement surgery and seeing if it delivers better or cheaper results than the existing models. CMMI can run the big randomized control trial between hip replacement payment models A and B, and if B turns out to cut costs by 40 percent while making no difference for patient outcomes, then Medicaid and Medicare can safely scale those interventions across their large patient populations. CMMI effectively does all the hard work of de-risking new payment and delivery models for CMS. And best of all, fewer people complain when an experiment fails to pay off because the whole point of experimentation (and therefore of CMMI) is to find out which new ways of doing things produce better outcomes and which ones do not.
An experimentation center at the National Institutes of Health, for example, could test out new funding mechanisms, like giving reviewers a “golden ticket” that they could use to ensure a particularly exciting proposal would receive funding. An experimentation center at the Department of Labor could pilot new models for retraining workers and provide an evidence base for future expansions. An experimentation center at the Department of Transportation could test out new schemes for delivering transit grants that drive cost reductions in public works. Thoughtful experimental design can help us to understand the underlying causal impact of possible reforms, and can also provide a way for bureaucracies to communicate the rationale for big policy changes in a manner that minimizes the appearance of arbitrary, unfair, or unaccountable decision-making.
In Revolt of the Public (2018), Martin Gurri discusses the phenomenon by which the legibility of the internet age has made the flaws of elites and public institutions more visible than they used to be (even if they are happening at the same rate). This pushes government agencies toward a kind of banal proceduralism, which often leaves the public unsatisfied, but is itself a reaction to public outcry over perceived unfairness. Consider, for instance, whether the “I’m using my discretion to give all my smart scientist friends grant funding” approach taken by Vannevar Bush, when he was running the Office of Scientific Research and Development (the precursor to the National Science Foundation) during World War II, would be applauded now.
Indeed, much of the institutional conservatism we see today stems from a reasonable desire to appear accountable and fair when deviating from an existing set of procedures, but which in practice creates a strong presumption for sticking with the status quo even when alternatives are desperately needed. Hence the importance of providing a scientific evidence base through experimentation to justify reforms (that often involve upsetting an established interest group) in a value-neutral language that public actors are well versed in and comfortable with. But to ensure that serious efforts at experimentation (and credible evaluation) are undertaken within a bureaucracy requires making it a discrete institutional goal.
One of the ongoing public reckonings relevant to the present discussion is why, exactly, the Centers for Disease Control and Prevention (CDC) failed so extensively at controlling and preventing disease during the Covid-19 pandemic. At nearly every turn, the CDC slow-walked the response and avoided taking proactive steps on issues like testing, while simultaneously boxing out other societal actors who were trying to fill the void.
The director of the agency, Rochelle Walensky, has acknowledged as much and specifically cited the tendency of the agency to act like an academic institution rather than an emergency responder. As Walensky said, the CDC was too focused on producing “data for publication” rather than “data for action.”
This is a startling contrast with the origins of the CDC in the 1940s as a quasi arm of the U.S. military, originally named the “Office of Malaria Control in War Areas,” which had a heavy operational focus in deploying medical countermeasures in the field. Over time, as malaria and other diseases were pushed back, the agency morphed into a more scholarly institution that attracted academics aiming to publish papers more than logistics and supply-chain professionals.
That’s not to say that scholarly work on disease spread is unhelpful. But an agency whose mission is fundamentally operational in nature—to take real action in the world to prevent the spread of disease—needs to find ways to maintain an operational culture over time. It perhaps should not be surprising that an agency run like a university would struggle when asked to turn into an emergency management operation overnight. To maintain this operational focus even when there’s not a pandemic to fight, the CDC should be practicing and operating systems in the real world—for instance by expanding a pathogen surveillance network in wastewater systems and airports, and by doing more work with international partners to proactively monitor and contain the spread of viruses around the world.
Indeed, the importance of maintaining an operational focus is more generally applicable for maintaining (or rebuilding) state capacity. There is a large literature in economics on the importance of learning by doing, i.e., the ability of firms to discover new efficiency improvements by repeating processes again and again. This repetition builds up tacit knowledge, which is embedded both in individual workers and in the larger organizational infrastructure. But this knowledge will erode if the communities of practice within which this process knowledge is mastered begin to atrophy. These are two sides of the same coin: practice generates learning by doing, but if we stop practicing, then our ability to do the activity at all can fade.
The state, too, can benefit from learning-by-doing loops through a focus on operational excellence. Sometimes that may require instantiating practices in the real world instead of in academic papers, as with the CDC. In other cases, it may require a shift in agency procedure and prioritization towards usability.
A basic truth of public service is that the government will do things that it has an office named for. Consider, for instance, the Census Bureau. The Census Bureau has an office of privacy; in fact, the bureau’s privacy procedures have recently become controversial because they’ve gone a step too far and made census data tangibly less useful in an effort to protect against hypothetical edge cases.
On the other hand, the Census Bureau has no office of usability. Compare this to an archetypal software company which prioritizes usability: half of this company might be focused on A/B testing small tweaks to the user interface to make its software slightly more intuitive or helpful. Yet emphasis on usability is largely missing from the public sector. A first step would be to have dedicated public offices within agencies that have a mandate to increase the usability of key services or products.
Not all agencies have an operational mission, of course. But for the ones that do, a focus on building and maintaining a community of practice by focusing on operational excellence should be a key strategy.
One analytical framework introduced by How to Make an Entrepreneurial State formalizes the different kinds of roles that innovation agencies can play within a larger ecosystem. Specifically, there are creators (like the National Labs), doers (like DARPA), funders (like NSF), intermediaries (like NIST), and rulers (like OSTP) that overlap with each other.
Moreover, a single organization can evolve over time from inhabiting one or multiple of these roles into others. Although not discussed in this sense in the book, NASA is a compelling example of an innovation agency that shifted from exclusively being a “doer”—directly manning and operating space exploration missions like Apollo—to a more complementary role as the market maker (“funder”) and shaper (“ruler”) of the commercial space sector today.
While NASA still directly operates a number of space missions, its arguably more impactful role today is as a facilitator of the nascent commercial space industry. SpaceX, in particular, has benefited tremendously from NASA’s clever use of a milestone-based payment model which let a scrappy start-up compete with incumbent space and aeronautics contractors in resupplying the International Space Station. In return, SpaceX has been able to massively drive down the costs of entering orbit by engineering a fully reusable orbital rocket that can be mass manufactured.
But all this was enabled by NASA’s willingness to reconceive of their core mission and intelligently leverage the flexible procurement authorization they had, called “Other Transaction Authority.” One can imagine a counterfactual scenario in which NASA instead insisted on doubling down on the consistently over-budget (but directly administered) Space Shuttle program or, alternatively, accepted dependence on the Russian Soyuz spacecraft.
Moreover, NASA continues to flex its muscles in its new “market maker” role with an intriguing contract to purchase lunar rocks from four different companies, conditional on them getting to the moon first. The idea here is twofold: first, to help jump-start a market by providing financial incentives to get to the moon; second (and perhaps more importantly), to establish some degree of international precedent for extraterrestrial property rights.
The agency will continue to wield substantial leverage in setting priorities for space exploration and commercialization over the coming decades. NASA is acting as the vanguard not only for space exploration but also for innovation agencies considering a similar transition from direct doer to market maker. Knowing if, how, and when other agencies should make a similar transition will need to be a core part of the entrepreneurial-state playbook going forward.
Another innovation agency using market-shaping financial mechanisms to its advantage is the Biomedical Advanced Research and Development Authority (barda). Barda is tasked with harnessing the private sector to fund defenses against high-risk, low-probability hazards that aren’t addressed by commercial markets. Using a broad suite of innovative procurement models, such as prizes, milestone-based awards and payments, advanced market commitments, and funding for university-affiliated research centers, barda is trying to proactively bootstrap and shape a market to fight pandemics, and to provide a defense against other chemical, biological, radiological, and nuclear incidents.
Essential to the success of this kind of operation is an appropriate grasp of the wide variety of financing mechanisms that exist. When is an innovation prize a better fit for a technical problem than a loan guarantee, advance market commitment, or a milestone payment? Mastering the nitty-gritty of government procurement best practices is not exactly a sexy policy topic, but it will be essential for building an entrepreneurial state.
How to Make an Entrepreneurial State asks the essential question for the twenty-first century. And the high-level frameworks and historical case studies it presents can help put us on the right track—even if this volume cannot function as a handbook.
The truth is that we don’t really know how to make an entrepreneurial state. There’s no single template that can be applied uniformly to rescue us from bureaucratic complacency. But the nature of entrepreneurship is to go out there and try new things. The strategies detailed above are in no way comprehensive, but they nonetheless have shown early signs of promise.
An underlying principle cutting across many of these strategies is a focus on creating new units within existing institutions that have a notable level of autonomy for the purpose of (a) providing internal competitive pressure, (b) de-risking and formalizing experimentation, (c) providing space for a community of practice and operational excellence, and/or (d) developing expertise in utilizing market-shaping mechanisms.
Often, a new office can achieve one or more of these goals at the same time. For instance, a new operational office in the CDC focused on proactive disease surveillance would certainly be attempting to achieve (a) and (c), but could perhaps involve (b) and (d) as well, depending on its structure.
It can be tempting for reformers faced with the obvious failure of our institutions to imagine the best path forward is a hard reset. But reforming our agencies won’t be as simple as turning them off and back on again.
Entrepreneurship often starts with small steps. The start-up that begins in the garage eventually grows up to take on the incumbent giant. Likewise, laying the seeds for effective institutional growth in small ways today can pay dividends down the road. And launching a new office or subagency is nearly always an easier political lift than creating a new agency from scratch.
This is certainly not to say that all issues of state capacity can be solved by creating new, autonomous offices, or that more ambitious reorganizations should never be considered. The use of market-shaping mechanisms, operational efficiencies, and experimentation should in no way be limited to particular offices and would ideally be embraced across all of government.
Our state capacity has been diminished over the years through the combination of conservative anti-statism and progressive proceduralism. To rebuild it will require a proactive vision of what we want the state to do and a clear-eyed understanding of the mechanisms needed to incentivize risk-taking.