Innovation has increasingly become a team activity, while inventors have become more concentrated in large firms. This column argues that these trends are closely connected. Team production creates genuine scale advantages in innovation, but can also lead the competitive economy to concentrate inventive activity too heavily in large firms because firms do not internalise knowledge spillovers.
Governments around the world devote substantial resources to R&D subsidies and tax credits, raising an increasingly important question: which firms should receive support? Large incumbent firms may be better positioned to innovate at scale, while smaller firms and entrants can generate new sources of growth and disproportionately large knowledge spillovers (Acemoglu et al. 2018, Akcigit and Kerr 2018, Takalo and Toivanen 2026). Recent work also cautions against using current firm size as a proxy for future innovative potential (Berlingieri et al. 2026).
One aspect of this debate has received less attention: the technology through which ideas are produced has itself changed. Innovation increasingly takes place in teams. When collaboration occurs primarily within firms, this shift can alter the returns to firm size and the allocation of inventors across firms.
In our recent research (Kim et al. 2026), we study the link between the rise of knowledge production in teams and the increasing concentration of inventors in large firms and analyse the implications for R&D policy.
In US patent data, the share of patents produced by teams of inventors increased from around 45% in 1976 to more than 80% by 2018, as shown in Figure 1. The same pattern emerges when patents are weighted by their market value or five-year forward citations.
Figure 1 Team patent shares
Firms play a central role in determining who collaborates with whom. Using new data on inventor employment histories, we find that nearly 90% of patenting collaborations occur among inventors working at the same firms. Firms provide a setting in which potential collaborators meet and form teams, so firm boundaries are an important constraint on team formation.
At the same time, inventor employment has become substantially more concentrated. Figure 2 shows that the employment share of inventors at large innovative firms has risen sharply since the late 1970s. Importantly, the increase in inventor employment concentration is roughly six times as large as the corresponding increase in overall employment concentration.
Figure 2 Inventor concentration and overall employment concentration
To explain these trends jointly and assess their implications, we develop a framework in which inventors can work individually or form teams with coworkers, and firms’ innovative activities generate knowledge spillovers to other firms. The key mechanism is intuitive. When a firm chooses not to adopt a team production model, each inventor’s productivity is independent of the number of other inventors in the firm. With team production, an additional inventor also expands the set of potential collaborators and increases the likelihood of finding a productive match. Larger firms therefore offer more opportunities to form productive teams, raising inventor productivity. This creates an endogenous scale advantage: as teams become more productive relative to solo invention with a larger pool of potential collaborators, large firms become increasingly attractive locations for innovative activity.
We estimate the model’s idea production function and knowledge spillovers using US patent data for two separate periods covering 1976–1999 and 2000–2018. Firms using team production exhibit higher returns to scale in inventor employment, and this scale advantage becomes substantially stronger over time. One possible interpretation is increasing knowledge specialization: as the knowledge required to innovate becomes more specialised, finding a complementary collaborator becomes more valuable, as emphasised by Jones (2009) and Freund (2025). At the same time, the estimated knowledge spillovers firms generate exhibit strongly decreasing returns to inventor employment.
In our model, these changes in the idea production function raise the team patent share, consistent with the data, because the allocation of inventors becomes more concentrated at large firms that organise inventors into teams. In our decomposition, the estimated changes in the idea production function account for the entire increase in inventor employment concentration.
The rise in inventor employment concentration driven by the rise of teams, combined with the observation that small firms generate knowledge spillovers that are disproportionate to their size (Chikis, Kleinman, and Prato 2025), provides a possible explanation for recent concerns about declining knowledge diffusion (Akcigit and Ates 2021, Jo and Kim 2026) and changing R&D productivity across large and small firms (Andrews et al. 2016, Cavenaile et al. 2025, Olmstead-Rumsey 2026).
A social planner, accounting for the effect of firm size on knowledge spillovers, also increases average firm size in response to the rise of teams, but by substantially less than in the competitive equilibrium. As a result, the output gains from reallocating inventors across firms rise more than fivefold, from 0.2% of output in the early period to 1.25% in the later period.
These results have direct implications for innovation policy. In our model, implementing the socially optimal allocation requires relatively greater R&D support for smaller firms, with an even stronger relative subsidy in the later period than in the earlier period. The reason is that firms’ private incentives capture a smaller fraction of the social value generated by an additional inventor through knowledge spillovers.
This perspective complements the current debate over the design of R&D policy. The relevant question is not simply whether governments should support ‘large’ or ‘small’ firms. Firm scale may reflect genuine technological advantages, particularly as innovation increasingly relies on teams. At the same time, changes in the organisation of innovation can endogenously alter the magnitude and distribution of knowledge spillovers across firms.
As innovation becomes more collaborative, policy therefore faces a balancing act: preserving the productivity gains from team-based innovation while supporting smaller innovative firms that contribute disproportionately to knowledge diffusion. Understanding how ideas are produced inside firms is thus central to designing R&D policy across firms.
Source : VOXeu
Direct evidence on the effects of financial sanctions remains limited. This column uses correspondent banking…
When could market stress justify European Central Bank intervention over French debt and what must…
The 2025-26 US tariffs have renewed the question of how much of a tariff reaches…
What matters more for sovereign credit ratings – today’s debt or tomorrow’s deficits? This column…
Populism may not only have contemporary but also long-term effects on the macroeconomy. Using data…
High public debt in several euro-area countries has renewed the debate over a central question:…