Innovation Ecosystems and the Complementarity Thesis: The Binding Constraints That Theory Left Unexplained
- Dr John H Howard
- 24 hours ago
- 9 min read
John H Howard, 17 March, 2026
My PhD examined the triple helix concept - Etzkowitz and Leydesdorff’s influential account of how universities, industry, and government interact to generate innovation. That was twenty-five years ago. The framework was genuinely illuminating: it identified institutional relationships that earlier linear models of innovation had obscured, and it gave analysts a vocabulary for describing the dynamics of knowledge economies that had not previously existed.
But something worried me throughout the research, and has continued over the two and a half decades since. The triple helix describes the actors and the relationships between them. It did not explain why the same configuration of actors produced transformative results in some locations and negligible returns in others. Silicon Valley had universities, industry, and government. So did every other advanced economy. The framework told you what to look for. It did not tell you what made the difference between places where it worked and places where it did not.
In the years that followed, I encountered successive frameworks that each added something real to the picture without quite closing that explanatory gap. Smart specialisation, cluster theory, startup-led innovation models, and place-based ecosystem thinking each illuminated a different facet of the same underlying phenomenon. Each was useful. Each left the same question unanswered.
The Complementarity Thesis
The Complementarity Thesis, developed in Making Sense of AI in 2026, published on 8 March, and drawing on the Handbook of Innovation Ecosystems, published in October 2025, may hold the analytical key that each of those frameworks was reaching for but never quite grasped.
The core claim is straightforward. The effects of any enabling capability, whether a technology, a policy instrument, or an institutional arrangement, do not depend on that capability alone but on the complements it combines with.
Where the necessary complements are absent, capabilities are unlikely to generate the returns they promise, however well-designed or generously funded they may be. The absence of a necessary complement is what makes it rate-limiting.
The metaphor comes from biochemistry, where the rate-limiting step in a reaction sequence is the slowest reaction in the chain: the one that constrains the overall rate of the process regardless of how abundant other inputs are. Increasing the concentration of other reactants does nothing to accelerate the system until the rate-limiting step is addressed.
In innovation ecosystems, a rate-limiting complement operates the same way. We can identify “leading complements” whose presence or absence disproportionately affects the ecosystem's overall productivity, regardless of how well other enabling conditions are in place.
They are “leading” in two senses: they tend to bind early and persistently across multiple stages of ecosystem development, and their resolution is typically a precondition for other complements to generate returns.
Research undertaken by the Acton Institute for Policy Research and Innovation suggests that the leading complements in innovation ecosystems are in the areas of social capital, talent density, patient capital, governance capacity, and market access. The scope of these rate-limiting complements is described in the table below.
Complement | Description |
Social capital | The accumulated stock of trust, reciprocity, and shared norms that enables knowledge exchange and collaborative risk-taking. The relational infrastructure of an ecosystem: invisible, slow to build, and slower still to recover once lost. |
Talent density | The concentration of people who combine technical depth with commercial acumen. The human capital that translates ideas, investment, and institutional support into innovative output, and the primary mechanism through which knowledge moves productively between organisations. |
Patient capital | Long-horizon investment willing to accept extended loss-making periods in exchange for the returns successful innovation eventually generates. The financial infrastructure that allows ventures to survive the gap between proof of concept and commercial viability — a gap that short-cycle investment consistently fails to bridge. |
Governance capacity | The institutional ability to make and implement strategic decisions across multiple actors, jurisdictions, and time horizons. The coordinative infrastructure that determines whether collective resources are directed coherently or dissipated in competing priorities. |
Market access | The availability of customers, domestic or international, willing and able to adopt innovative products at commercially meaningful scale. The demand-side condition that converts innovative capability into sustainable economic activity — and the complement whose absence most reliably explains why technically excellent ecosystems fail to produce globally competitive firms. |
Note: Each leading complement is necessary but not sufficient on its own. The diagnostic task is to identify which is currently rate-limiting in a specific ecosystem at a specific stage of development.
The rate-limiting complement chain is represented in Figure 1 below.
Figure 1: The Rate-Limiting Complement Chain
© Acton Institute for Policy Research and Innovation, 2026
There are many other “supporting” complements, such as the presence of anchor tenants, the quality of intermediary organisations, the depth of research and development capacity in proximate universities, scale-up and testing facilities, specialist legal and financial services familiar with innovation transactions, technology transfer offices, education and training in innovation management, and the regulatory conditions that determine whether productive experimentation is permitted or inadvertently discouraged.
The nature of these connections is illustrated in Figure 2. It shows how supporting complements reinforce leading complements in what we increasingly understand as complex dynamic systems.
Figure 2: Leading and Supporting Complements in Innovation Ecosystems
© Acton Institute for Policy Research and Innovation, 2026
The connections in Figure 2 indicate how a supporting complement reinforces multiple leading complements. The “lattice effect” reflects the interdependent, non-linear character of the system: intervening at any node produces effects that propagate across the whole. The diagnostic task is to identify which nodes are rate-limiting at a particular point in time and are likely to be in the future.
What makes leading complements analytically useful is precisely that they are not equally binding at every stage of ecosystem development.
As suggested in Figure 3 below, the diagnostic task is not to assemble all of them simultaneously, which is neither feasible nor necessary, but to identify which ones are rate-limiting and binding at a particular juncture in ecosystem development and which are likely to be the most appropriate for the design of instruments that will generate value-creating returns.
Figure 3: Diagnostic Matrix: Complements by Ecosystem Stage
© Acton Institute for Policy Research and Innovation, 2026
Addressing a non-binding complement, however worthy the intervention, is unlikely to accelerate the system. The history of innovation policy is substantially a history of applying well-designed instruments to the wrong constraint.
Figure 3 also suggests that no single complement binds across all stages of ecosystem development: social capital is the critical constraint in early-stage ecosystems, while governance capacity and market access become binding only as an ecosystem matures. It also offers a simple explanation for why so much innovation policy disappoints.
Governments may continue to apply instruments that are well-suited to early-stage ecosystems, with the result that the ecosystem never matures. For example, an ecosystem that continues with heavy investment in social capital programs, is likely to find itself well-networked, well-populated with startups, and persistently unable to scale.
As ecosystems mature, relationships between people tend to move from the informal to more professional and business like as financial and other major resource commitments become more prominent. But it is still accepted wisdom that “people do business with people they trust”. In parallel, governance moves from networks and collaborations to more structured (rules driven) and strategic relationships as asset values build and prudential and accountability demands intensify.
At the same time, Governments routinely apply instruments that are well-suited to more mature ecosystems such as statutory governance frameworks, market development programs, international linkage strategies, to ecosystems that are still constrained by social capital deficits and thin patient capital. The instruments are not wrong, but they may be premature.
The question is never whether an ecosystem needs better social capital, talent pipelines, patient capital, governance, frameworks, or market access. It always needs all of these.
The question is which of them is currently the binding constraint, the rate-limiting complement whose absence and design is limiting the returns on everything else. To urban planners, architects, economists, business leaders, and governments, these returns will be ultimately reflected in increased employment, sales of goods and services, exports, returns on investment, and productivity growth.
The diagnostic matrix can also be used in different industrial contexts, from biotechnology, software, manufacturing, agriculture, energy and environment through to arts and creative practice. The stages from the early stage through to maturity are likely to differ markedly across sectors.
The Triple Helix in a New Light
Returning to the triple helix with this lens makes the framework considerably more precise. The three bilateral relationships it identifies, between universities and industry, universities and government, and industry and government, each require their own distinct set of connective complements before they can generate useful innovation output. The presence of the three institutional actors is necessary. It is nowhere near sufficient.
The influence of connective complements in the triple helix framework is summarised in Figure 4.
Figure 4: The Triple Helix with Connective Complements
© Acton Institute for Policy Research and Innovation, 2026
The university-industry interface is the most studied and the most consistently misunderstood. The binding complement is not proximity, nor even formal partnership agreements or technology transfer offices, though these matter. It is trust: the accumulated relational capital that allows knowledge to move across organisational boundaries without being captured, distorted, or ignored. Formal mechanisms can facilitate trust but cannot substitute for its absence.
Three connective complements give trust its practical expression: personnel exchange between universities and firms, which builds shared vocabulary that formal collaboration cannot manufacture; joint governance of intellectual property agreed before research generates value rather than negotiated after; and shared investment in research infrastructure, which aligns incentives in ways that contract research alone rarely achieves.
The university-government interface has a different binding complement: alignment of purpose across institutions operating on fundamentally different time horizons. Universities plan in decades; governments plan in electoral cycles. The connective complements that bridge this gap include mission-oriented research programs with genuine time horizons, funding structures that reward commercial relevance without distorting research independence, and intermediary organisations capable of translating between the two institutional logics.
Australia's Cooperative Research Centre model, in place since 1992, has been the most sustained attempt to build this infrastructure. Its record is instructive: strong where mission alignment was genuine, weaker where the partnership was assembled to access funding rather than to pursue a shared objective.
The industry-government interface turns on regulatory intelligence: the capacity of government agencies to understand the technologies and business models they are regulating well enough to make calibrated judgements, rather than applying precautionary frameworks designed for earlier technology generations. Government can also act as a demanding customer for innovative products and an investor in enabling infrastructure that the private sector will not provide at the required scale or time horizon.
What the triple helix has always implied, but rarely made explicit, is that these three relationships are not independent. Each bilateral interface is shaped by what both partners need from the third institutional actor, and each is, in this sense, a rate-limiting complement for the other two. A weakness in any one of them constrains the productivity of the whole system, regardless of how well the other two are functioning.
Place-based innovation policy that addresses only one or two of the three interfaces reliably underperforms its own expectations.
Concluding Comment
What the Complementarity Thesis adds to two and a half decades of wrestling with the triple helix is a diagnostic discipline that the framework has always lacked. Rather than asking whether the three institutional actors are present and interacting, it asks which connective complement is currently binding in each bilateral relationship, and what that implies for where investment and policy attention should be directed.
The arrival of GenAI sharpens this diagnostic challenge considerably. AI does not dissolve the connective complement problem; it reconfigures it.
In the university-industry interface, AI is generating new tensions around intellectual property, data ownership, and the attribution of commercially valuable outputs that existing trust frameworks were not designed to handle.
In the university-government interface, the challenge of mission alignment intensifies when the technology is moving faster than policy frameworks can track.
In the industry-government interface, the regulatory intelligence complement becomes more binding, not less, as AI systems contribute to decisions that were previously the exclusive domain of accountable human judgement.
A UTS research program is currently examining more than 20 global innovation districts to test a direct proposition arising from this framework: that ecosystem complement quality, more than technology adoption intensity, explains variation in AI productivity outcomes across locations.
About the Author
John H. Howard is Director of the Acton Institute for Policy Research and Innovation and Honorary Visiting Professor at the University of Technology Sydney’s Institute for Public Policy and Governance. He is currently managing the UTS research project examining how innovation ecosystem quality explains AI productivity outcomes across global innovation districts. He is the author of The Handbook of Innovation Ecosystems (Acton Institute Publishing, 2025) and Making Sense of AI in 2026 (Acton Institute Publishing, 2026).
Sources
Comments