Innovation Economics

Lead Summary

Innovation economics is the body of theory examining how technological novelty is generated, diffused, and transformed into economic growth. It departs from neoclassical economics in a fundamental way: rather than treating technology as an exogenous force delivered from outside the economic system, it locates innovation at the heart of competitive dynamics and long-run development. Growth, in this view, is not primarily the result of more capital or more labour but of continuous qualitative transformation — new products, new production methods, new organizational forms — displacing what came before.

The field draws on three overlapping traditions: Schumpeterian evolutionary economics, which models capitalism as a system of creative destruction powered by entrepreneurial innovation; endogenous growth theory, which formalizes the role of non-rival knowledge and R&D spillovers in sustaining long-run growth; and the National Innovation Systems literature, which examines how countries' institutional configurations shape their capacity to innovate. Running across all three is a shared preoccupation with dynamics over statics, with disequilibrium over equilibrium, and with the role of institutions, market power, and the state in shaping who innovates and who benefits.

Origins and Background

Schumpeter's break with statics

The intellectual starting point for innovation economics is Joseph Schumpeter's insistence that the essence of capitalism is not price competition but revolutionary transformation. He drew a sharp conceptual line between invention and innovation: inventions are technological discoveries, while innovations are the successful economic implementation and commercialization of those discoveries. The entrepreneur is not primarily the inventor but the innovator — the person who organizes production, financing, and markets to bring new combinations into the economy, bearing the risk of doing so. Economic progress depends not on scientific discovery alone but on entrepreneurial implementation.

Schumpeter also rejected static competition analysis — fixating on price levels and concentration ratios — in favour of dynamic competition emphasizing the threat of entry, new technologies, and product differentiation. A firm with a monopoly today may be displaced tomorrow by a superior entrant. That threat, not the number of current competitors, is what disciplines markets and drives progress.

The concept that crystallised this dynamic view was creative destruction: capitalism as a continuous revolutionary transformation of the economic structure, perpetually destroying old forms and creating new ones. This is not merely a feature of capitalism; it is, for Schumpeter, its defining mechanism.

Neoclassical absence

The mainstream neoclassical tradition proved unable to accommodate this insight. Its equilibrium framework requires complete markets, perfect information, and rational actors — conditions under which genuine entrepreneurship, bearing radical uncertainty, cannot be modelled. Neoclassical equilibrium theory systematically omits the entrepreneur despite entrepreneurship being central to actual economic processes. This represents a foundational gap between equilibrium theory and historical economic reality.

The Austrian school had its own heterodox response: the entrepreneur as alert discoverer, competition as an ongoing process rather than an equilibrium state. But it was the evolutionary economics of Nelson and Winter and the endogenous growth theory of Romer and Aghion-Howitt that gave the Schumpeterian tradition its formal academic footing.

Core Concepts

Non-rival knowledge and increasing returns

The foundation of modern innovation economics is a seemingly simple observation: ideas are non-rivalrous. One person's use of a mathematical theorem, a programming language, or a manufacturing design does not prevent anyone else from using the same thing simultaneously. This characteristic means that once an idea is created, it can be widely disseminated at minimal cost.

When non-rival ideas are combined with rival inputs — capital and labour — the economy exhibits increasing returns to scale. If rival inputs are doubled along with ideas, total production more than doubles. This mathematical property provides an endogenous mechanism for sustained economic growth, distinguishing endogenous growth models from neoclassical exogenous models where growth eventually runs out as capital accumulation faces diminishing returns.

R&D spillovers and the case for intervention

R&D investments by individual firms generate positive externalities that benefit the entire economy. When a firm invests in research and development, part of the resulting technological knowledge spills over to other firms without the original firm being compensated. This occurs because knowledge is non-excludable. These spillovers represent a fundamental market failure: unregulated markets produce less technological change than is socially optimal.

The policy implication is direct: subsidies for R&D and investments in education increase the steady-state growth rate by correcting for positive externalities that private actors underinvest in. More broadly, policies that promote openness, competition, and technological change generate higher growth. This is the canonical economic rationale for public research funding and education investment.

The monopoly-innovation trade-off

In endogenous growth models, firms investing in R&D to create new blueprints face high fixed costs that must be recouped. Temporary monopoly power — gained through patents, trade secrets, or first-mover advantage — allows firms to charge prices above marginal cost, generating the profits needed to finance R&D. Without this market power, competitors could immediately copy innovations at minimal cost, eliminating the incentive to invest.

This creates a fundamental trade-off: some degree of monopoly power is necessary to generate sufficient innovation incentives, yet monopoly pricing creates allocative inefficiency in the short run. The result is that unregulated markets produce some innovation but less than is socially optimal.

Temporary monopolies may therefore be necessary for and conducive to innovation, challenging the static welfare analysis that treats monopoly as unambiguously detrimental. Monopoly rents provide both the incentive and the means for large firms to invest in long-term research. The elimination of monopoly profits in pursuit of static efficiency could reduce the incentives for breakthrough innovations that drive dynamic growth. The "monopoly" rents are the price that society pays for progress.

Static versus dynamic efficiency

A direct consequence is the tension between static and dynamic efficiency. Static efficiency means minimizing deadweight loss and maximizing current consumer surplus through perfect competition. Dynamic efficiency means achieving the socially optimal level of long-run R&D investment. Schumpeterian competition pursues the latter, often at the cost of the former: it does not seek price competition but fundamentally overturns the existing economic order, which may involve higher prices in the short run but generates innovation gains in the long run.

Contemporary antitrust and competition policy faces a growing consensus that it must balance these competing objectives rather than pursue static efficiency alone — a debate that has become urgent in the context of digital platform markets.

Endogenous Growth Theory

Romer's formalisation

Paul Romer's 1990 paper "Endogenous Technological Change" provided the first rigorous endogenous growth model, earning him the 2018 Nobel Prize in Economic Sciences "for integrating technological innovations into macroeconomic analysis." The prize recognition explicitly acknowledged that endogenous growth theory provides essential frameworks for understanding long-term economic growth and for analyzing technological solutions to climate and environmental problems.

Romer's model treats the stock of knowledge as a productive input that grows endogenously through deliberate R&D. Because knowledge is non-rival, a researcher's use of existing knowledge does not deplete it — enabling the compounding accumulation that sustains growth indefinitely.

The Aghion-Howitt quality ladder

Aghion and Howitt's 1992 paper formalized Schumpeter's intuitions into an endogenous growth model. Their "quality ladder" framework shows how growth is driven by vertical innovations under monopolistic competition: firms innovate to create superior products that displace existing technologies, earning temporary monopoly profits before the next innovator displaces them in turn. The model endogenizes creative destruction by showing how the prospect of future innovations discourages current research — by threatening the rents from current innovations — creating a trade-off between static and dynamic efficiency that drives sustained growth. This work earned Aghion and Howitt the 2025 Nobel Prize in Economic Sciences.

The scale-effects problem

Early Romer-style models predict that steady-state growth rates are proportional to the number of researchers, implying that larger populations should achieve higher growth. Empirical evidence contradicts this prediction. The United States experienced substantial increases in scientists and engineers in R&D between 1950 and 1988, yet total factor productivity growth remained relatively constant. Countries with larger populations have not systematically grown faster.

More broadly, empirical support for endogenous growth theory is limited and mixed. Time-series studies show no consistent positive relationship between R&D personnel growth and TFP growth. Endogenous growth models also fail to explain conditional convergence documented in cross-country data — the phenomenon that countries with similar fundamentals converge to similar income levels regardless of initial conditions. These "scale effects problems" prompted important theoretical refinements, but the basic empirical relationship between R&D stock and growth remains difficult to pin down.

Schumpeterian Regimes: Mark I and Mark II

The distinction between entrepreneurial (Mark I) and routinized (Mark II) patterns of innovation has become one of the most productive taxonomies in the field. The labels were formalised post-hoc by Phillips (1971) to contrast two visions embedded in Schumpeter's own evolving work.

Mark I captures Schumpeter's early view (The Theory of Economic Development, 1934): individual entrepreneurs as primary drivers of discontinuous innovation. This regime is characterized by low entry barriers, reliance on external learning and spillovers, stochastic innovation outcomes, and high firm turbulence with frequent entry and exit. Technological competition takes the form of creative destruction — successful innovators replace incumbents.

Mark II captures his later view (Capitalism, Socialism and Democracy, 1942): large incumbent corporations becoming dominant innovators, rendering individual entrepreneurs economically redundant. This regime is characterized by high entry barriers, concentration of innovation in few large firms, cumulative technological capabilities, and predictable innovation outcomes following established R&D routines. Technological competition takes the form of "creative accumulation" — incumbents continuously strengthen dominant positions through incremental improvement.

These modes are not fixed to particular industries but conditional on technological opportunity conditions. High appropriability and high cumulativeness systematically favour Mark II patterns: when incumbents can retain rents through patents or tacit knowledge, and when current innovations necessarily build on accumulated prior knowledge, innovative advantage concentrates in established firms. Low appropriability and low cumulativeness favour Mark I by enabling independent entry.

Evolutionary Economics: Nelson and Winter

Nelson and Winter's 1982 work An Evolutionary Theory of Economic Change formally modelled Schumpeterian innovation dynamics at the firm and sectoral level, replacing neoclassical assumptions of profit maximization and equilibrium with evolutionary processes modelled on biological natural selection.

Routines as genes

Organizational routines function analogously to genes — repetitive patterns of activity that store and transmit an organization's operational knowledge, serving as the primary replicators in evolutionary economic change. Routines enable firms to operate predictably and carry forward learned solutions to recurring problems, making them the fundamental units of selection in evolutionary economics.

Variation and selection

Nelson and Winter establish a dual-mechanism framework: variation occurs through firm-level search and routine innovation, while selection operates through competitive market forces. More profitable firms expand market share; less profitable competitors decline and exit — without requiring that any firm consciously optimizes. This variation-selection separation mirrors biological evolution and operationalizes how economic systems generate novelty while simultaneously testing viability through market discipline.

Local search and competence destruction

Established firms search locally for innovative solutions, exploring possibilities close to their existing knowledge base. This emerges rationally from the high costs and tacit knowledge requirements of searching far from current competencies. As a consequence, firms' innovation trajectories remain constrained by historical capabilities.

When a paradigm shift arrives, this becomes a liability: technological discontinuities can be competence-destroying. Major innovations render established firms' existing routines and capabilities obsolete. Incumbent firms face severe difficulty acquiring new competencies optimized for prior technologies, explaining why radical innovations often displace market leaders despite their resources and experience.

Nelson and Winter also distinguish between science-based regimes (knowledge flows from external scientific research into commercial application) and cumulative regimes (narrow knowledge bases built through continuous internal learning). Cumulative regimes are stable and predictable but potentially path-dependent; science-based regimes are more volatile and potentially disruptive. Most industries exhibit hybrid characteristics.

Techno-Economic Paradigms and Long Waves

Carlota Perez extended the Schumpeterian framework into a comprehensive theory of long-wave economic cycles organized around major technological revolutions. A Techno-Economic Paradigm (TEP) is a complex collective learning process that articulates a dynamic mental model of best economic, technological, and organizational practices during the period when a specific technological revolution is being adopted. When adopted widely, it becomes the foundational "common sense" for organizing activities and structuring institutions across diverse sectors.

Perez traces five major technological revolutions in capitalist history, each representing a distinct surge spanning roughly fifty years:

1. The Industrial Revolution and mechanization
2. Steam power and railways
3. Steel, electricity, and heavy engineering
4. Oil, automobiles, and mass production
5. Information technology and telecommunications

Each surge follows a structural pattern of installation and deployment phases separated by upheaval. During the installation phase, financial capital leads adoption through speculative investment and creative destruction, characterized by Schumpeterian turbulence. During the deployment phase, state support and production capital take the lead, and the technologies diffuse throughout society creating widespread prosperity.

Financial bubbles are inherent features of the installation phase — not market failures but natural mechanisms through which finance mobilizes capital for radical uncertainty. The deployment phase, by contrast, requires state activism to achieve the full growth potential of the paradigm and to ensure that benefits diffuse equitably beyond financial centres. The transition from installation to deployment thus depends not merely on technical maturation but on socio-institutional adaptation — emergence of state policies appropriate to the deployment phase.

The fifty-year periodization is presented as an empirical observation of historical capitalism rather than a strict natural law, echoing and operationalizing Kondratieff wave theory while grounding the cycle specifically in the time required for full paradigm emergence, installation, deployment, and maturation.

National Innovation Systems

Definition and origins

The National Innovation Systems (NIS) concept was formulated in the late 1980s and early 1990s through collaborative work by Christopher Freeman, Bengt-Åke Lundvall, and Richard Nelson. Freeman defined NIS as the network of institutions in public and private sectors whose activities and interactions initiate, import, modify, and diffuse new technologies. Lundvall emphasized organizations and institutions involved in searching and exploring, including all aspects of economic structure affecting learning. Nelson focused on the set of institutions whose interactions determine the innovative performance of national firms. The framework crystallised in three foundational texts: Freeman's Technology Policy and Economic Performance (1987), Lundvall's National Systems of Innovation (1992), and Nelson's National Innovation Systems: A Comparative Analysis (1993).

The NIS framework serves two complementary objectives: explaining international differences in countries' capacity to innovate, and deriving policy recommendations for enhancing firms' innovative activities within national contexts. These dual objectives — explanatory and prescriptive — distinguish NIS from purely theoretical economics and position it as a framework for both scholarly analysis and policy action.

The Triple Helix

The Triple Helix model, developed by Etzkowitz and Leydesdorff in the 1990s, proposes that innovation emerges from interactions among three institutional spheres: universities (knowledge generation), industry (commercialization and production), and government (policy and oversight). As interactions between these spheres intensify, each evolves to adopt characteristics of the others, creating hybrid institutional forms — technology transfer offices, science parks, venture capital firms, incubators, and accelerators. The model emphasizes innovation in the innovation process itself: the creation of mechanisms through which these three sectors can collaborate.

Industrial clusters and agglomeration

Industrial clusters and geographic agglomeration generate substantial innovation advantages through multiple complementary mechanisms: knowledge spillovers between proximate firms, specialized labour markets concentrating workers with specific skills, and specialized suppliers reducing transaction costs. Geographic concentration enables the tacit knowledge diffusion, face-to-face interaction, and informal information networks that depend on proximity.

Silicon Valley exemplifies these self-reinforcing dynamics, combining venture capital, human capital, university ties, government support, industrial structure, and support services — creating advantages difficult for other regions to replicate despite decades of deliberate policy efforts.

NIS limitations in developing countries

The NIS concept was originally developed by and for high-income industrialized countries, creating substantial challenges for its direct application in low- and middle-income contexts. Developing countries have weaker science and technology activities, less formalized university-industry linkages, fewer venture capital mechanisms, limited R&D infrastructure, and different state-market relationships. These are not merely quantitative gaps but structural differences requiring substantial conceptual and methodological adaptation rather than direct transfer.

Effective adaptation requires: reconceptualising innovation to include social and organizational innovations alongside technological ones; incorporating informal economy and grassroots innovation rather than focusing only on formal R&D; adjusting measurement indicators appropriate to institutional realities; recognizing that different development stages require different institutional emphases; and examining innovation inequalities within countries alongside international comparisons.

Contemporary research identifies multiple distinct catch-up pathways: a "balanced catching-up" cluster (Ireland, Spain, India, Russia); an "imbalanced catching-up" cluster concentrating resources on specific high-capability sectors (South Korea, Taiwan, China); and "trapped NIS" economies stuck at middle-income stages. Some resource-based economies like Chile and Malaysia demonstrate pathways through developing high-value-added resource-sector goods — contradicting assumptions that manufacturing-led catch-up is the only viable path.

South Korea and Taiwan exemplify this institutional diversity: South Korea through a high-debt, chaebol-dominated structure combining heavy government direction with large-scale integrated investment; Taiwan through an SME-based networked structure blended into global production networks through original design manufacturing. Both achieved rapid technological catch-up through fundamentally different institutional configurations.

The Entrepreneurial State

Mariana Mazzucato's entrepreneurial state thesis challenges the dominant narrative attributing technological progress solely to private enterprise. Her argument: the state acts as a market-creator and lead risk-taker in innovation, funding the highest-risk, most uncertain phases of research before the private sector is willing. Rather than merely fixing market failures, the state proactively shapes and creates new technological opportunities before their commercial potential is recognized by business communities.

The evidence marshalled includes:

• The US National Science Foundation funded the foundational search algorithm underlying Google's search engine — a case of public research directly enabling private commercial success.
• The Internet originated from state investment in ARPANET before private firms commercialised it.
• Apple popularized government-created technologies — GPS navigation, touch-screen technology, voice recognition — into the modern smartphone.
• The NIH funded foundational molecular biology techniques that private pharmaceutical companies subsequently commercialized. Private VC entered only after the state reduced uncertainty and demonstrated scientific and commercial viability.
• Green technology and renewable energy sectors demonstrate ongoing state entrepreneurialism: state funds basic and applied research in renewable technologies, develops infrastructure standards, and creates demand through policy before private markets engage substantially.

While value creation is collective — involving state researchers, entrepreneurs, workers, and public institutions — its distribution is not. The state socializes risks through public R&D but does not capture proportional returns.

The political implication: while value creation is collective, its distribution is not. The state socialises risks through public R&D investment but does not capture proportional returns, while private corporations and shareholders extract disproportionate value from innovations built on public foundations. Mazzucato advocates mission-oriented innovation policy aimed at socialization not only of risks but of rewards — through equity stakes, licensing revenues, or other mechanisms for public value capture.

Venture capital's exit-driven model (3–5 year horizon toward IPO or acquisition) is, on this account, structurally limiting for long-run innovation in sectors like biotechnology, where the highest-value research unfolds over decades, not investment cycles.

Innovation and Inequality

The distributional ambiguity

Innovation has ambiguous effects on income inequality. In the short run, innovation increases inequality at the top of the distribution because innovators earn rents from new technologies. However, when innovation comes from new market entrants rather than incumbents, it simultaneously creates social mobility and lifts economy-wide productivity.

Innovation's effect on inequality depends critically on whether it comes from new entrants or incumbent firms. Entrant innovation generates positive effects on social mobility and reduces concentration of wealth held by existing elites, whereas incumbent innovation does not significantly affect social mobility. This explains why innovation increases top income inequality through rent generation while simultaneously increasing social mobility through creative destruction — making it largely uncorrelated with broader Gini measures.

Incumbent entrenchment

Yesterday's innovators become entrenched incumbents that use lobbying and political influence to prevent future innovation and new entry, capturing rents and limiting the creative destruction mechanism. The positive correlation between innovativeness and social mobility is significantly dampened in states and regions with higher lobbying intensity. The decline of US productivity growth, together with increases in concentration and rents since the early 2000s, illustrates incumbent entrenchment effects.

Skill-biased technological change

The spread of computer technology from the 1970s significantly increased relative demand for educated workers, generating substantial wage inequality. Industries that adopted computer technology more rapidly experienced larger wage premiums for skilled workers. Computer adoption accounts for approximately 30–50% of the growth in relative demand for skilled labour.

Automation can also generate major wage inequality while producing only modest productivity gains. The task-based framework shows that displacement effects from automation reduce labour's share of value-added — explaining why many workers experience wage stagnation or decline even during periods of rapid innovation and productivity growth.

Superstar firms

In technologically dynamic industries, winner-takes-all patterns have emerged. Approximately one-third of the growth in US household income inequality since 1980 can be explained by the compensation gap between superstar firms — characterised by extreme market concentration, high markups, and dominant market shares — and other employers. These firms develop competitive advantages through technological improvement, augmenting sales at much higher rates than workforce expansion. But the same concentration stifles innovation among laggard firms and new entrants with difficulty raising capital.

Distance to frontier

The inequality-innovation relationship differs by development stage. In economies at the technological frontier, higher inequality may amplify growth by providing investment capital for frontier innovation. In economies further from the frontier, inequality can impede growth through creative destruction because wealthy households satisfy demand for quality goods through imports rather than supporting local innovation and firm entry. This asymmetry suggests that a single policy approach to managing the inequality-growth-innovation relationship is inappropriate across different development contexts.

Creative destruction and productivity

Empirical research on US manufacturing confirms Schumpeter's core mechanism: entry and exit of firms accounts for over 50% of ten-year productivity growth between 1977 and 1987. New plants enter at initially lower efficiency than incumbents but catch up through learning and selection. Old, inefficient capital is destroyed; productive new capital is created.

Creative destruction depends fundamentally on low barriers to entry. When entry costs and expansion costs are high, the reallocation mechanism weakens: inefficient firms cannot exit and productive new firms cannot expand. In economies with high institutional or regulatory barriers to entry, the creative destruction process is substantially weakened, slowing productivity growth.

Intellectual Property: Incentive or Obstacle?

The patent paradox

While patent theory posits that protection provides incentives by allowing inventors to recoup fixed costs, the general equilibrium effects can be negative. Monopoly pricing restricts follow-on innovation, access to knowledge inputs, and incremental competition. Empirical evidence does not support the claim that patents are the primary driver of productivity growth: the number of patents awarded shows no correlation with measured productivity gains.

Historical evidence reinforces this: 89% of British innovations at the 1851 World's Fair were never patented. Inventors employ alternative appropriation mechanisms — trade secrecy, first-mover advantage, complementary assets — whenever patents are ineffective or impractical.

Patent effectiveness varies significantly across industries. Patents are most effective in pharmaceuticals and chemicals, where high fixed development costs and ease of reverse engineering make them essential for recouping investment. In electronics, digital industries, and services, their direct effect is more modest. One-size-fits-all patent regimes may therefore provide excessive protection in some sectors while being inadequate in others.

Patent trolls and thickets

Non-practicing entities (patent trolls) cause substantial harm to innovation. After receiving an initial NPE lawsuit, targeted firms reduce their innovation in those technology areas and shift R&D to avoid legal exposure. Non-target firms in related technology areas also move their innovation away from targeted areas, creating chilling effects across entire technology clusters. NPE lawsuits caused approximately half a trillion dollars of lost wealth to defendants between 1990 and 2010.

Open source as an alternative

Open source represents a viable alternative to patent-based incentive systems for cumulative innovation, particularly in software. Research modelling R&D competition shows that open source equilibria emerge when research is highly complementary, R&D costs are large, and firms are sufficiently different from each other. While open source generates smaller per-firm rewards compared to patent monopolies, it enables all participants to use the technology and attracts new entrants, inducing faster overall innovation pace.

Global South patent gaps

Only approximately 2% of global patenting originates from Latin America and the Caribbean combined, with similarly low percentages from Africa. This disparity reflects measurement bias as much as innovation capacity: conventional indicators systematically undercount learning-based, incremental, and locally-adapted innovations throughout the Global South. Emerging economies including India, Vietnam, Egypt, and South Africa have increased innovation capabilities in ICTs, biotechnology, and engineering — but conventional metrics create a distorted global picture where innovation appears concentrated in the Global North.

Technology Diffusion and Catch-Up Growth

The advantages of backwardness

Gerschenkron's "advantages of backwardness" theory posits that economically backward countries possess a structural advantage in rapid industrialisation: they can adopt mature technologies at lower cost than the original innovators incurred, enabling compressed industrialisation timelines. The technology gap between backward and advanced economies, rather than constituting an insurmountable disadvantage, enables latecomer firms to implement proven technologies without bearing full R&D costs.

However, diffusion is far from automatic. Technology diffusion patterns across developing countries are highly uneven, with imported technologies frequently failing to reach even 25% population penetration. Technology gaps close quickly when cost differences are small, but persist for decades when labour costs and implementation expenses differ substantially between developed and developing contexts.

Absorptive capacity

Cohen and Levinthal's (1990) absorptive capacity — a firm's ability to recognise the value of new external knowledge, assimilate it, and apply it commercially — is the critical mediator of technology adoption. Absorptive capacity is primarily a function of prior related knowledge accumulated through R&D investments, and the relationship is path-dependent: early investments in specific technical areas enable future capability development, while early gaps may foreclose future opportunities in those areas.

The middle-income trap

Countries that successfully reach middle-income levels through technology diffusion often face growth slowdowns and fail to achieve high-income status. Escaping this "middle-income trap" requires a fundamental transition: from adopting and diffusing foreign technology to generating original, domestically developed technology — a qualitatively different capability requirement. Countries become caught between rapid technological change in advanced economies and wage-based competition from lower-income countries.

Controversies and Debates

Does innovation cause or reduce inequality?

The empirical relationship between innovation and inequality is genuinely contested. Innovation increases top income inequality through rent generation, simultaneously increases social mobility through entrant-led creative destruction, and is largely uncorrelated with overall Gini measures — a pattern that is coherent theoretically but difficult to communicate politically. The relationship further differs by distance from the technological frontier: near the frontier, inequality may amplify growth; far from the frontier, inequality can dampen it.

Large firm versus small firm innovation

Whether innovation is better generated by large established firms or small entrepreneurial ventures remains empirically unresolved. Schumpeter's own view shifted across his career on this question — from emphasizing individual entrepreneurs in 1911 to large R&D-capable firms in 1942. Contemporary evidence shows both large and small firms innovate, in different sectors and at different stages. The superstar firm phenomenon suggests that large digitally-enabled firms can concentrate both innovation and market power simultaneously, complicating both hypotheses.

The entrepreneurial state: state as investor or crowding-out force?

Mazzucato's thesis that the state is a lead risk-taker and market creator has generated significant scholarly debate. Critics note that government-backed failures (Solyndra, various industrial policy disasters) receive less attention than successes, and that selection bias shapes the evidentiary record. Defenders note that VC similarly uses profits from 1 of 10 successes to cover 9 failures — and the state has not been allowed to reap returns commensurate with its risk-bearing, creating systematic asymmetry.

Frontier technologies and global inequality

Every major technological wave from mechanization through computerization has deepened relative inequality between advanced and developing economies unless accompanied by deliberate policy. Frontier technologies — AI, advanced robotics, precision agriculture, gene-editing — carry the same risk. Developed countries concentrate both frontier technology innovation and early adoption advantages. Without policy intervention favouring technology diffusion and local adaptation, frontier technologies will amplify inequality rather than support inclusive development.